3D Histology Imaging and Image Reconstruction Methods

(1)

Jonna Jokela

3D HISTOLOGY IMAGING AND IMAGE RECONSTRUCTION METHODS

Faculty of Medicine and Health Technology

Bachelor’s Thesis

April 2021

(2)

ABSTRACT

Jonna Jokela: 3D histology and image reconstruction methods Bachelor’s thesis

Tampere University

Faculty of Medicine and Health Technology April 2021

Histology is commonly used in a wide range of applications. It studies the microscopic structure of different biological tissues, usually in two dimensions. However, there are different three- dimensional (3D) methods available for studying the volumetric structure of a tissue in scale of micrometres.

In this thesis, the principles of three 3D histology methods are reviewed and applications of them in medical field are presented. The aim is to summarize previous studies of micro-computed tomography (micro-CT), selective plane illumination microscopy (SPIM) and serial block face scanning electron microscopy (SBF-SEM) and give an overview of the applications of each method. Since each of the methods have different physical basis and equipment, all of them have also certain characteristics and specific tissue types that are most easily imaged with them. In addition to general overview, this thesis aims to study if some of the methods have applications in breast cancer research and whether they could be useful when breast cancer is diagnosed.

Based on the literature, each one of the methods is useful in their application areas and provide new information of tissue structure compared to conventional histology. The most suitable method has to be selected based on the tissue type and the target of study. The developing technology of micro-CT enables imaging of various tissue types with constantly increasing resolution. It is suitable for both hard and soft tissue types, such as bone, brain, lung or intestinal tissue. In addition, micro-CT has many promising applications in the field of breast cancer research and diagnosis. For instance, the intraoperative determination of tumour margin status would be possible by micro-CT. SPIM is particularly suitable for embryology because it allows, with certain modifications, the continuous scanning of the specimen. This allows studying, for instance, how metastases are formed in cancer. The most recent method, SBF-SEM, is especially useful in imaging of neural tissue, and it allows also visualization of intracellular structures in three dimensions. As the methods develop and become more automatic and accurate, the amount of useful applications might increase.

Keywords: three-dimensional imaging, micro-CT, SPIM, SBF-SEM, breast cancer, histology

The originality of this thesis has been checked using the Turnitin OriginalityCheck service.

(3)

TIIVISTELMÄ

Jonna Jokela: 3D-histologia ja kuvan muodostamisen menetelmät Kandidaatintyö

Tampereen yliopisto

Lääketieteen ja terveysteknologian tiedekunta Huhtikuu 2021

Histologia on menetelmä, jolla on monia käyttökohteita lääketieteessä. Se tutkii kudoksen mik- roskooppista rakennetta, yleensä kahdessa ulottuvuudessa (2D). On kuitenkin olemassa useita erilaisia menetelmiä, joiden avulla voidaan tutkia kudoksen tilavuudellista rakennetta kolmessa ulottuvuudessa (3D) mikroskooppisella tasolla.

Tässä kandidaatintyössä käsitellään kolmea erilaista 3D-histologian menetelmää ja niiden sovelluskohteita lääketieteessä. Työn tavoitteena on perehtyä kirjallisuusselvityksen pohjalta mikro- tietokonetomografian (engl. Micro-computed tomography, micro-CT), selektiivisen tasova- laisumikroskopian (engl. Selective plane illumination microscopy, SPIM) ja lohkopinnan sarjata- sovalaisumikroskopian (engl. Serial block face scanning electron microscopy, SBF-SEM) erityis- piirteisiin ja sovelluskohteisiin. Työssä muodostetaan katsaus jokaisen menetelmän perustaan ja kuvantamiseen vaadittavaan laitteistoon, sekä arvioidaan, millaisille kudostyypeille ja millaisiin sovelluskohteisiin menetelmät soveltuvat parhaiten. Lisäksi työssä tutkitaan, onko menetelmiä sovellettu rintasyövän tutkimiseen ja olisivatko ne hyödyllisiä rintasyövän diagnosoinnissa.

Kirjallisuusanalyysiin perustuen, jokainen tutkituista menetelmistä on hyödyllinen omissa so- velluskohteissaan ja mahdollistaa uudenlaisen tiedon hankkimisen kudoksen rakenteesta verrat- tuna perinteiseen histologiaan. Näytteeseen ja sovelluskohteeseen perustuen on valittava sopivin menetelmä. Micro-CT:n kehittyvä teknologia mahdollistaa monenlaisien kudostyyppien kuvanta- misen jatkuvasti paremmalla resoluutiolla. Se sopii sekä kovien kudosten, kuten luun, että peh- mytkudosten, kuten aivojen, keuhkojen ja suoliston, kuvantamiseen. Lisäksi micro-CT:llä on monia lupaavia sovelluskohteita rintasyövän tutkimisessa ja diagnosoinnissa, kuten poistetun syö- päkudoksen reuna-alueiden kuvantamisessa leikkauksen aikana. SPIM on erityisen hyvä mene- telmä alkiotutkimukseen, sillä se mahdollistaa tietyin muokkauksin näytteen jatkuvan kuvantami- sen, ja täten sen avulla voidaan saada lisää tietoa esimerkiksi metastaasien muodostumisesta syövässä. Uusin menetelmistä, SBF-SEM, sopii erityisesti esimerkiksi hermokudoksen kuvantamiseen, sekä solujen sisäisten rakenteiden kuvantamiseen kolmessa ulottuvuudessa. Tulevai- suudessa menetelmiä on mahdollista kehittää entistä helppokäyttöisemmiksi ja tarkemmiksi, jol- loin hyödyllisten sovelluskohteiden määrä mahdollisesti kasvaa entisestään.

Avainsanat: kolmiulotteinen kuvantaminen, micro-CT, SPIM, SBF-SEM, rintasyöpä, histologia

Tämän julkaisun alkuperäisyys on tarkastettu Turnitin OriginalityCheck –ohjelmalla.

(4)

LIST OF SYMBOLS AND ABBREVIATIONS

2D Two-dimensional

3D Three-dimensional

CCD Charge couple device

CT Computed tomography

DCIS Ductal carcinoma in situ

ER Endoplasmic reticulum

FLIM Fluorescence lifetime imaging GFP Green fluorescent protein

HER2 Human epidermal growth factor 2 H&E Hematoxylin and eosin

Mars-SPIM Multiangle-resolved subvoxel selective plane illumination microscopy Micro-CT Micro-computed tomography

MRI Magnetic resonance imaging

NA Numerical aperture

SEM Scanning electron microscopy

SBF-SEM Serial block face scanning electron microscopy SPIM Selective plane illumination microscopy

TNM Tumour classification method

UV Ultraviolet

(6)

1. INTRODUCTION

Histology, the study of biological tissues at microscopic level, has many applications in medical field, from analysing the cellular structure of different tissue types to histopathological research of diseased tissue. Mostly, the tissue is sliced into very thin sections and examined through a conventional light microscope. However, this only allows observing the internal structure of the specimen in two dimensions while the volumetric structure is disregarded. Despite of its popularity, conventional histology has many disadvantages.

The tissue preparation is time-consuming and requires plenty of different equipment and reagents. In addition, the tissue slicing may break some of the internal structures of the sample and usually every part of the tissue does not get to be viewed under a microscope which may compromise the coverage of its analysis. (Glaser et al. 2017)

There are various imaging methods available that aim for deeper understanding of the microscopic structure of a biological tissue. The aim of this thesis is to introduce three imaging methods that produce visualization of the sample in three dimensions in microscopic level. In addition, the aim is to estimate whether some of these methods would be applicable in imaging breast tissue or lymph nodes when breast cancer is detected. The three-dimensional (3D) imaging methods selected are micro-computed tomography (micro-CT), selective plane illumination microscopy (SPIM) and serial block face scanning electron microscopy (SBF-SEM). They have very different technological basis and different targets of application which this thesis will present in the form of a literary review.

This thesis aims to provide an overview of the principles of micro-CT, SPIM and SBF- SEM and presents how they can be applied to different areas of histological research.

The methods will be compared to each other and to conventional histology. Additionally, this thesis will analyse the future prospects of the 3D histology methods both in general and in breast cancer research and detection.

The first method discussed in the thesis, micro-CT, shares its working principle with X- ray-based computed tomography (CT). CT is normally used in hospitals to detect diseased or injured areas of the body, for instance, in cancer detection or bone fracture detection. The main difference between the methods is that micro-CT possess higher resolution power which allows visualization of the microscopical structure of different organs or tissues. Micro-CT has become popular in 3D histology applications because it is

(7)

non-invasive, non-destructive and normally the sample do not require further preparation besides the application of contrast agents in non-mineralized tissue. (du Plessis et al.

2017) The second method, SPIM, is based on a phenomenon called fluorescence. It is developed for the needs of developmental biology since imaging of whole, living embryo has appeared to be a very challenging object of study. It has many advantages, such as the non-invasiveness and the high resolution, and nowadays, it has been applied to wide range of research fields. (Greger et al. 2007) The third one, SBF-SEM, combines an electron microscope and a diamond knife which together allow the visualization of cells and cell organelles in three dimensions. It has the most applications in the field of neu- roimaging, but new subjects are discovered continuously. (Peddie & Collinson 2014) Breast cancer is one of the most common cancers in the world. More than one in every eight women are diagnosed with breast cancer during their lifetime, which is approximately 12,9 percent of women, according to National Cancer Institute of USA (2020). In detection of breast cancer, the most commonly used methods are mammography, ultrasound, magnetic resonance imaging (MRI), CT and histopathology. All of them have particular drawbacks, such as the high expertise and experience of histopathologist that is required for diagnosis. The aim of this thesis is to study micro-CT, SPIM and SBF-SEM as possible methods for breast cancer detection and diagnosis, and discover whether they could overcome the limitations of conventional methods.

The thesis is structured as follows. Chapter 2 includes the theoretical and technical back- ground of micro-CT, SPIM and SBF-SEM and introduces the imaging setup of each.

Chapter 3 explains the basic methods in preparation of a histological specimen and provides information about how the preparation is performed both in conventional histology and in 3D histology methods. In chapter 4, the recent applications of each method in the field of medicine are discussed. Chapter 5 contains general information about breast cancer histology. In addition, the potential of 3D methods in breast cancer detection and research is discovered and summarized in chapter 5. Finally, chapter 6 summarizes the findings of this thesis and compares the methods based on recent research.

(8)

2. METHODS FOR 3D HISTOLOGY IMAGING

In this chapter, micro-computed tomography, selective plane illumination microscopy and serial block face imaging are introduced and principles of them are gone through.

2.1 The principle of micro-computed tomography

Micro-computed tomography is an imaging method that can be used to investigate the structure of tissue with resolution of 1 μm (Tamminen et al. 2020). It is based on electromagnetic X-rays that are produced by an X-ray tube. The X-rays interact with the sample and penetrate through it, and finally a detector collects the rays and forms a 2D projection image. Different tissue types absorb and scatter photons differently, depending on several factors, such as the tissue density and atomic number. Hence, tissues can be dis- tinguished from each other since they appear with different shades of grey in the image.

In addition, the characteristics of the X-ray beam also affect the achieved contrast. As the sample is scanned from numerous directions, the 3D structure of the sample can be reconstructed from the projection images using different algorithms. (du Plessis et al.

2017)

The X-rays are generated by accelerating electrons in a vacuum tube. The cathode of the X-ray tube emits electrons when it is heated to a high temperature. These electrons are then accelerated towards the anode due to high voltage between the anode and the cathode of the X-ray tube. (Huda 2016) For biological samples, the recommended tube- voltage is from 30 to 100 kV (du Plessis et al. 2017). The electrons interact with the atoms of the anode material in many different ways. The interaction with nuclear electric field cause the electrons to decelerate and change their direction which causes so-called Bremsstrahlung photons to be generated. The Bremsstrahlung method is the most used technique in computed tomography applications. (Huda 2016)

In micro-CT scanner, the sample is placed between the X-ray source and the detector, as shown in figure 1. In medical CT-scanners the imaged patient stays stationary while the scanning device moves around the patient, but in micro-CT the imaged sample is rotating between the X-ray source and the detector (du Plessis et al. 2017). The produced X-rays can interact with the imaged sample in various ways. They can scatter via coherent scattering or Compton scattering or be absorbed by the atoms of the matter via the photoelectric effect. The rest, usually less than 1% of the X-ray photons, will pene-

(9)

trate through the sample and be collected on the detector. There are many different detector types available, such as combinations of a scintillator and a charge-couple device (CCD). The scintillator is a device that emits visible photons when X-ray photons excite the atoms of it, in other words it is based on fluorescence. The CCD converts the light photons to electric charge, forming a digital image. (Huda 2016)

Figure 1. The principle of micro-CT. The sample is irradiated with X-rays produced by an x-ray tube source after which the formed projection images are registered by a detector. The rotation stage allows imaging from many angles. (Modified

from Farncombe & Iniewski 2017)

The total scanning time is dependent on the time that is needed for taking single projection image. It can vary between detectors due to the differences in sensitivity and dy- namic range. In addition, the acquisition time of a projection image affects the contrast and quality of the image. The number of different angles the sample is imaged at de- pends on the sample size and the desired resolution. The bigger the sample is and the higher the desired magnification is, the more steps are required to achieve a convenient 3D reconstruction. (du Plessis et al. 2017)

There are many different 3D reconstruction software available which map each imaged voxel using a Feldkamp filtered back-projection algorithm (Feldkamp et al. 1984). After computational processing, the volumetric data is visualized as a surface view or as a volumetric 3D visualization. In addition, many microstructural characteristics can be cal- culated from the 3D dataset. (du Plessis et al. 2017)

While the projection images are observed, different artifacts may be present. If the sample moves during the imaging, it may cause blurring in 3D reconstruction. To achieve high-quality images, the resolution, the number of imaging angles and the X-ray voltage need to be chosen properly. For example, too high voltage may cause poor contrast.

The most common artifact in micro-CT is called beam hardening. It is caused by dense parts of the sample since the low-energy X-rays are absorbed differently in those areas.

Beam hardening can be prevented by using filters either in front of the sample or between

(10)

the sample and the detector. In addition, the X-ray beam has lower intensity on the edges, resulting in artifacts on the edges of the image. This is also preventable by using only the middle of the detection area. The reconstruction process can help to remove some of the artifacts but it can also cause e.g. double edges when incorrect parameters are used. (du Plessis et al. 2017)

2.2 The principle of selective plane illumination microscopy

Selective plane illumination microscopy (SPIM), which is also called single plane illumination microscopy or light sheet microscopy, is based on fluorescence. SPIM is a note- worthy method for 3D histology imaging since it is minimally phototoxic and achieves high resolutions. Fluorescence is based on an ability of certain molecules to excite to higher energy level and emit light with a longer wavelength when they are illuminated with electromagnetic radiation. The wavelength of illuminating radiation is usually in ultraviolet (UV) region or in the blue region of visible light, in other words in region from 360 nm to 400 nm. Substances can be divided into fluorophores and fluorochromes. The former have natural ability to fluoresce, for example vitamin A. Hence, such biological components can be detected as they are. The latter are different dyes and chemicals which are added to tissue to produce fluorescence because not every molecule has the property naturally. Such molecules require to be stained in order to be detected. (Ban- croft et al. 2019, p. 36)

In SPIM, a regular fluorescence microscope is placed above an imaged specimen. A particular characteristic of SPIM is that the specimen is illuminated from the side using a collimated laser and a cylindrical lens, as presented in figure 2. The cylindrical lens fo- cuses the laser to the focal plane of the tube lens. This causes excitation and fluorescence only in fluorophores that are in a thin volume around the focal plane. This reduces photodamage and phototoxic effects in the other parts of the sample since only the detected part is illuminated. The photodamage is also reduced because each plane of the imaged specimen is illuminated only once during the detection. (Huisken et al. 2004;

Greger et al. 2007) In addition, accurately limited illumination extends the operating time of the fluorophore because it reduces the amount of photobleaching. Bleaching occurs because every molecule has a limited number of times it can excite and emit fluorescence light, causing eventually fading and ending of the fluorescence phenomenon as the sample is illuminated (Lichtman & Conchello 2005). In SPIM, the illuminated slice of the sample emits a fluorescence signal which is detected by a CCD of a camera. Addi- tionally, the measurement unit contains a filter that blocks irrelevant light caused by the

(11)

laser or fluorophores that fluorescence in a wrong wavelength. (Greger et al. 2007; Ban- croft et al. 2019)

Figure 2. The principle of SPIM. The sample is illuminated from the side by a fo- cused light sheet. A laser and a cylindrical lens are used to form the light sheet which causes fluorescence in desired volume. The detection system contains a regular fluorescence microscope and a movement unit. (Modified from Greger

et al. 2007)

The movement unit of SPIM can move and rotate the imaged sample relative to the detection and illumination systems. In that way the imaging system can acquire a stack of images by illuminating the sample slice by slice. By performing the imaging from various directions and recombining the stacks of images, SPIM can produce structural information of the sample in three dimensions. (Greger at al. 2007) This method is called multiview SPIM. An advantage of SPIM is that there is no need for tomographic reconstruction but the unprocessed 3D stack of images acquired by the camera is usually usable as it is (Huisken et al. 2004; Funane et al. 2018). In addition, SPIM has the ability to perform live imaging with high spatial and temporal resolution which enables for example imaging live embryos. (Greger et al. 2007; Fu et al. 2016) This method is called fluorescence lifetime imaging (FLIM) (Funane et al. 2018).

The resolution of SPIM is defined by the detection microscope and the thickness of the light sheet formed by the laser. The lateral resolution is the same as in regular wide-field systems because it is dependent on the objective lens of the detector. The thickness of the light sheet determines the axial resolution. The thinner the light sheet is, the smaller field of view can be examined and vice versa. The thickness itself is determined by the

(12)

numerical aperture (NA) of the illumination object. In general, the resolution reached is under 1 μm in both directions. (Greger et al. 2007)

2.3 The principle of serial block face scanning electron micros- copy

Serial block face imaging or serial block face scanning electron microscopy (SBF-SEM) is an imaging method introduced by Denk and Horstmann in 2004. The method starts with preparation of an imaged specimen by embedding it in a resin block. A scanning electron microscope (SEM) is used to image the surface of the sample by sending an electron beam that scatters back from the block surface. The back-scattered electrons are detected and a digital image of the surface is produced based on the intensity differences in returned electrons. In addition to the SEM system, the chamber contains a diamond knife which is called a microtome, that cuts an ultrathin slice from the sample and removes it. The slice is usually in range of nanometres. After that, the block surface is scanned with electrons and imaged again. The process is repeated until a desired volume has been scanned. As a result, the method produces hundreds or thousands of images which can be used to reconstruct the 3D structure of the sample with an image analysis software. The reconstruction process of SBF-SEM is more effortless than in methods that image the removed slices and align them afterwards because the sample block stays stationary during the process and obtained images are automatically in line.

The imaging cycle is presented in figure 3. (Hughes et al. 2013; Cocks et al. 2018)

(13)

Figure 3. The principle of SBF-SEM. The sample is lifted up to a desired cutting height, a microtome slices the top of the sample and a scanning electron micro-

scope detects the electrons scattered from the block face. (Hughes et al. 2013) The preparation stage is important when SBF-SEM is used and it takes hours or even days to prepare the tissue for imaging. The process includes staining the tissue with heavy metal (for instance, lead, osmium or uranium), washing the tissue numerous times and embedding the sample with resin. Staining is performed in order to improve the contrast especially in membranes. There are various staining protocols available for that purpose, e.g. zinc iodide method. (Wilke et al. 2013) Tissue preparation is covered in chapter 3 in more detail.

Since the tissue is sliced between every image, the sample embedding affects the image quality by changing the hardness and cutting properties of the tissue. If the material is too soft for cutting, the axial resolution may decline since the system manages to cut only thicker slices. In addition, the staining can cause charging and conductivity within the sample. The sample charging can cause problems in resolution because the formed charges can cause undesired reflection of the electron beam. At the same time, the sample conductivity usually improves the resolution because with a conductive sample, the charge can be conducted to the ground. (Hughes et al. 2013, Smith & Starborg 2019) Another difficulty is that the metal atoms inside the specimen can cause knife marks to emerge on the surface of the sample when the atoms adhere to the knife edge. Knife marks affect the roughness of the imaged surface which can change the scattering angles the electrons have, causing challenges to the imaging procedure. (Hashimoto et al.

2016)

(14)

3. PREPARATION OF HISTOLOGICAL SPECIMEN

Histology can be defined as a study of biological tissues in microscopic level. It combines study of organs, different tissue types and cells. In order to analyse those, proper tissue preparation is needed. Tissue preparation includes usually following stages: fixation, dehydration, clearing, infiltration and embedding of the tissue. Selected methods for each stage depend on the characteristics of the tissue and the cells, together with the object of study. (Bancroft et al. 2019, p. 74) It is important to understand the most common stages of tissue preparation since the same principles are valid both in conventional histology and 3D reconstruction methods, and chosen chemicals affect the achieved image quality.

3.1 Fixation of the sample

The first step of histological preparation, called fixation, can be either a physical or a chemical method. A commonality between different methods is the purpose of them. The fixative should maintain the micro-architecture of the tissue by preventing enzymatic activity, dissolution of cellular components and activity of pathogens. In other words, the most important purpose of a fixative is preventing the destruction process of a tissue since the tissue may be stored even for years. Choosing a proper fixation method is very important since it may modify, for example, the quality of staining later on the preparation process. In fact, fixative impacts every stage of the preparation process. (Bancroft et al.

2019, pp. 40-41)

Physical fixation is usually done by using heat, microwaves or freeze-drying. Chemical methods can be divided into three categories: coagulant methods, cross-linking methods or combination of different methods. In coagulant methods, the most used compounds include alcohols and acetone which coagulate the cellular proteins. Unfortunately, these methods are not worthwhile for analysing the ultrastructure of cells since they precipitate the cytoplasm. Cross-linking methods form supporting chemical bonds between proteins and they include compounds such as glutaraldehyde, chloral hydrate, various metal salts and formaldehyde. Formaldehyde is the most used cross-linking method in histological applications and it especially affects amino acids, proteins and lipids. For example, the electron microscope in SPIM system usually requires fixing the tissue with cross-linking fixative, such as compound of glutaraldehyde and formaldehyde, to prevent damage in membranes within the cells. As this fixative is a combination of two ingredients, it can be

(15)

also considered as a compound fixative. Also, other compound fixatives which have formaldehyde as other component are commonly used in histopathological diagnostics. For instance, breast tissue should be usually fixed with 10 % neutral buffered formalin for 6- 8 hours if the tissue is sliced during the research, as in SBF-SEM. (Bancroft et al. 2019, pp. 40 – 50, p. 55)

3.2 Dehydration, clearing and infiltration of the sample

The second stage of preparation, dehydration, is based on graded alcohols, such as ethanol and isopropanol, which remove free water in the tissue, while saving the water integrated to molecules of the cell. Dehydration is needed because it makes the tissue responsive to clearing agents and infiltration wax. Clearing, which is the third stage of preparation, is used to remove the alcohols added in dehydration stage and act as a lipid solvent. The clearants usually contain xylene, a chemical which has the ability to act as a tissue dryer and has detrimental effects to health. Hence, other similar chemicals, such as d-limonene or aliphatic hydrocarbons, can substitute for xylene nowadays. (Bancroft et al. 2019, pp. 74-74)

After three stages of processing, the tissue is ready for infiltration, the fourth stage of the preparation. Since the tissue is sliced in many different applications, for example in SBF- SEM, the infiltration is of great significance. There are various waxes available for different tissue types, mainly selected based on the tissue density and the desired slice thickness. The most used one, paraffin wax, has many advantages, like the low price and the compatibility with different tissue types and staining methods. It is also usable for embedding after the infiltration, in which a solid tissue block which allows slicing is formed.

The main purpose of paraffin is to provide mechanical support during the slicing. In addition, agar gel, gelatine, celloidin and resin are also used as infiltrators. For instance, resin operates often as the infiltrator and the embedding material in electron microscopy applications, such as SBF-SEM in which the tissue block is sliced with an ultramicrotome between the images. (Bancroft et al. 2019, pp. 74-77) In addition, freezing the tissue is also an option when a microtome is used, for example, during a surgery to perform a more justified diagnosis (Bancroft et al. 2019 p. 88). However, cutting frozen or embedded sections may destroy the tissue structure and cause artefacts to the image. For instance, breast tissue does not often freeze without any difficulty which may cause problems in an intraoperative diagnosis. (Glaser et al. 2017)

(16)

3.3 Hematoxylin and eosin staining

After the tissue is processed properly and cut into thick slices, the sections are stained in order to differentiate different cellular structures in microscopic examination. In most cases of tissue staining in conventional histology, a combination of hematoxylin and eosin (H&E) is used since it has many advantages from high quality to ease of use. The H&E staining is often automatized in laboratories as it is so widely used. Also, it is commonly used in histopathological diagnosis, for example, when a pathologist examines a histological section in order to diagnose cancer.

Hematoxylin is used to stain the cell nuclei blue. It can be used separately, but usually it is combined with eosin which stains the cytoplasm and tissue fibres pink. Hematein, the oxidization product of the logwood extract, is a natural stain that causes the blue colour of hematoxylin. Hematoxylins are classified into different categories according to mordant that is needed for hematein to be able to act as a stain. The mordant is a metallic salt, for example, aluminium or tungsten, that forms a complex with hematein. The most used mordant is aluminium which forms so called alum hematoxylin. Eosin is the most used stain with alum hematoxylin, and the combination of them results in blue-pink staining of the cells. There are various types of eosin available, the most used one being eosin Y. If some specific structure has to be differentiated more accurately, there are numerous specific staining methods available. (Bancroft et al. 2019, pp. 126 - 127)

3.4 Sample preparation for micro-CT, SPIM and SBF-SEM

The steps explained above are mostly used in conventional histology, but the same principles apply also to micro-CT, SPIM and SBF-SEM.

To perform micro-CT-imaging, the specimen does not have to undergo a long preparation process but it can be imaged as it is. However, the preparation has direct effects on the image quality, whereupon the specimen is usually processed somehow. The most crucial factor that affects the image quality is avoidance of the sample movement which may occur during the rotational scanning process. Hence, the sample must be attached securely to the sample holder of the equipment. Also, when biological tissues are imaged, dehydration and drying of the specimen must be taken into account, for example, by immersing the sample in the suitable liquid during the scanning. In addition, staining can further improve the image quality. It is especially important in non-mineralized structures. For instance, substances with high atomic numbers, such as osmium, gold, iodine or lead, are suitable for micro-CT staining since they absorb X-rays and improve the visualization of different structures. (Mizutani & Suzuki 2011; du Plessis et al. 2017) For

(17)

instance, in study by Virta et al. (2020), the group used iodine-based staining to improve the contrast as they produced 3D visualizations of intestinal samples.

SPIM requires embedding the sample in gel, for example, in agarose. That way the sample stays still on the screen, allowing imaging a living specimen, such as embryos, for several days since the agarose gel is biocompatible. Usually, the embedded sample is surrounded with aqueous medium that provides more ideal imaging since the refractive indexes of the embedding material and the liquid are almost identical and the outermost lens of the detection system can be immersed in the liquid. To perform SPIM, the sample must be fluorescent which can be achieved by using fluorescent proteins or stains. There are numerous different stains available that bind to specific molecules inside the cells, making them visible by illumination. In addition, there are three different fluorescent proteins which are green, yellow and red. They are commonly used in biological applications of fluorescence microscopy. For instance, in embryo imaging the red fluorescent protein can work as a light emitter in a specific wavelength. (Greger et al. 2007)

Technology of SBF-SEM requires carrying out a few of the preparation stages carefully.

At first, the sample is usually treated with formalin or glutaraldehyde for fixation, after which the tissue is stained with heavy metal. As mentioned before, the heavy metal staining is crucial because the metal atoms work as reflectors of the electron beam, improving the contrast. However, the preparation process is very time-consuming since the staining substance needs a long time, even a week, to penetrate through the tissue. To overcome this problem and to speed up the process, the sample is usually chopped into small pieces, in range of millimetres. At the end, the specimen is embedded in resin to form a block suitable for cutting with the diamond knife. (Smith & Starborg 2019)

(18)

4. APPLICATIONS OF 3D METHODS

In this chapter, the current application areas of micro-CT, SPIM and SBF-SEM are presented in addition to which the advantages and limitations of each method are studied based on observations made in different research frames. Applications of micro-CT are divided into two separate chapters since it is widely used in specific breast cancer applications in addition to other medical purposes. Breast cancer associated applications of micro-CT are presented in chapter 5.

4.1 Applications of micro-CT

Micro-CT is a versatile imaging method as can be noticed from the amount of different branch of sciences that have gotten benefit from the use of it in research. As it was initially developed for other than medical applications, it has been applied to material sciences, geosciences, palaeontology and multiple other research areas. In medical field, micro-CT has been used in imaging various tissue types, such as bone tissue and dental tissue, and numerous types of soft tissue.

An example of the usefulness of micro-CT in dental research is a study by Tomaszewska et al. (2018) where they studied the structure of the root canal of a molar tooth. They were able to reconstruct the root canal and study the size and orientation of it in detail, and they gained information that differed from 2-dimensional knowledge since the structure seemed less complicated than in previously published studies. In similar way, micro- CT has been applied to other dental structures, such as detection of caries. (Boca et al.

2017) Also, in bone structure imaging micro-CT has proven to be a very useful method since it can be applied to osteoporosis diagnosis, osteoarthritis modelling and analysis of overall structure of the bone. (Xu et al. 2013, Chang et al. 2021)

Micro-CT can also be applied to imaging different soft tissues. For instance, Senter-Za- pata et al. (2016) used micro-CT in imaging brains of a mouse and both uterus and lung tissue of a human, and discovered the effects of conventional histology preparation. The tissue slices shrunk 19–61% during embedding which was detected by imaging them with micro-CT before and after it. They expected that to be a result of tissue breakage or dehydration during the embedding process. Also, the staining caused the sample shrink- age to be smaller than without it. In addition to the volume differences, they studied the 3D histology reconstruction from serial slices which was compared with the reconstruction by micro-CT. They stated that micro-CT was very useful as a guidance method for

(19)

tissue slicing and that it could be used in common histology methods to first image the block with micro-CT and after that slice it, knowing the tissue’s morphology and quality.

Using micro-CT as an auxiliary method for histology could give really useful data about the consequences of tissue processing and make analysing histology slices more precise.

Another example of micro-CT in soft tissue imaging is a study by Kirschner et al. (2016) where the group imaged brain tumours of mice with micro-CT and clinical CT device, and compared the results to conventional histology slices. There were not major differences between the detected tumour volumes of micro-CT and clinical CT, but there appeared to be a volume change of 33 % in histological fixation and embedding. The results also support the previously mentioned usefulness of micro-CT as a supportive method for conventional histology. In the study, the dose of micro-CT appeared to be five times higher than the dose of the clinical scanner which may be considered when animal mod- els are studied in the future.

Similar comparison of micro-CT and histology was made when Virta et al. (2020) analysed intestinal samples in order to improve diagnostics of celiac disease. An advantage of micro-CT was the digital, endlessly repeatable sectioning of the sample since histological diagnostics suffer from demanding determination of suitable cutting angle. In the research, micro-CT was able to diagnose 2 of the 8 biopsies that histological assay could not detect as diseased to be that sort. Although resolution of micro-CT could not overcome the one of conventional histology and further studies are needed to confirm the results, micro-CT could be a potential method for celiac diagnostics and treatment trials.

In addition, micro-CT has been applied to the research field of medical implants. For instance, micro-CT can be used in detection of osseointegration of dental implants in three dimensions. Song et al. (2021) performed micro-CT for dogs that had a dental screw in their mouth and compared the results with paraffin embedded histological slices.

The results based on two-dimensional (2D) slices of micro-CT and histology were very similar, but there were some differences between the histological results and the results of 3D micro-CT in the detected volumes. Hence, they suggest the correlation between 3D micro-CT and histological detection to be studied further to prevent overestimations.

Another example of implant imaging is a study by Teymouri et al. (2011) where the group used micro-CT and histology as a basis for comparison in evaluating the placement of a cochlear implant by clinical CT after the surgery. The results of each method were compatible which is encouraging as clinical CT is commonly used in such applications. In addition, the artifact caused by metallic parts of the implant in reconstruction of micro-

(20)

CT could be removed by setting a intensity threshold value and discovering the area of the device from the data sheet of it.

The technology of micro-CT is evolving continuously, leading to higher achievable resolution. In a study by Tamminen et al. (2020), the attained voxel size was 1,1 μm which enabled studying cellular structures of thousands of cells and detection of a subcellular phenomenon. They managed to examine nuclei shapes of cells in the culture with high accuracy when compared to optical imaging methods. They also stated that micro-CT would be applied to imaging both cellular structures such as junctions between different cells, or components of extracellular matrix. However, micro-CT is not suitable for studying structures smaller than the micro-scale since the voxel size is not adequate for that purpose. For instance, nano-computed tomography (nano-CT) has higher resolution and hence is more suitable for studying smaller structures.

In conclusion, micro-CT is a versatile method with applications in several fields of sci- ence. As it is a constantly developing method, the amount of different branches of sci- ence to which micro-CT is applied might be even increased in the future.

4.2 Applications of SPIM

SPIM is commonly used for studying development of intact, living embryos. Embryos are usually in size of a few millimetres which is a relatively large size in optical microscopy.

When imaging this large specimen, the light scatters and is absorbed into the tissue which lowers the resolution and the quality of images. However, if the sample is rotated during the imaging, the desired resolution can be achieved. For example, Huisken et al.

(2004) imaged live embryos of teleost fish Medaka and fruit fly Drosphila melanogaster with conventional SPIM-setup and green fluorescent protein (GFP). They succeeded in imaging the whole Medaka embryo in 3D with resolution higher than 6 μm in depth of 500 μm. In the study, they rotated the sample into four different orientations in order to achieve 3D reconstruction with high resolution. Besides the whole embryo, they imaged the heart of the Medaka embryo and were able to reconstruct the internal structure of it, including ventricle and atrium, and the heartbeat as a video. The sample rotation resulted in four image stacks that were usable even without further processing.

In addition to the large size of the specimen, the opacity of the specimen also affects the image quality commonly in optical microscopy. To show the performance of SPIM, Huisken et al. (2004) also imaged a fruit fly embryo (Drosophila melanogaster) which appears to be more opaque than other embryos, and they examined that SPIM was capable of reconstructing the embryo even without imaging it from many directions. The

(21)

embryo also developed normally despite the high amount of fluorophore excitations during the study which indicates the lower phototoxicity of SPIM compared to other fluorescence methods. In addition, in a study by Krzic et al. (2012), an embryo of Drosophila melanogaster was imaged with a bit different setup of SPIM, in which there were two illumination and two detection lenses which imaged the specimen in two axes perpendicular to each other. This is called multiview SPIM which does not require rotation of the specimen, resulting in less phototoxicity, higher resolution and faster imaging since rotation can cause artefacts to the images because of the sample movement. The imaging time was very short, only 20 seconds for the whole embryo. As they also studied the effects of slow specimen rotation in order to improve the quality further, scanning of the sample from eight directions was achieved in time four times shorter than with conventional SPIM due to the imaging from multiple directions at time. On the whole, SPIM is an excellent method for developmental biology since it can achieve resolution below cellular level, it is usable in real time and an embryo can be scanned on the whole while it still stays alive during the process. The same principles might be usable also for other specimen in the scale up to a few millimetres.

The second application of SPIM is imaging the central nervous system which has been studied my multiple authors. For instance, Panier et al. (2013) studied the use of SPIM in imaging activity of a Zebrafish brain and achieved promising results. They succeeded imaging several regions of the brain in three dimensions by moving the sample into five different heights relative to the light sheet and recording the calcium activity of the neurons. However, during a single recording, they imaged 25 000 neurons which is not enough for formulating a model of the neural network. The sample movement turned out to have some disadvantages, such as the undesired stimulation of the neural system and extension of the imaging time since the sample has to be stationary in order to achieve high quality. In addition, the system did only provide single-cell resolution in the proximal parts of the brain because of the light sheet shape. However, Panier et al.

(2013) state that creation of a brain atlas will be possible with SPIM in the future.

Another example of examining the neural tissue is a study by Funane et al. (2018) where a combination of SPIM and FLIM was produced for lifetime imaging of a cleared mouse brain by injecting GFP carrying viruses into the brain before the removal process. Alt- hough they managed to collect 3D lifetime data with axial resolution of 9,9 μm and lateral resolution of 7 μm, the imaging time appeared to be quite long, approximately 100 minutes for single plane. SPIM-FLIM is an informative method when different neurologi- cal diseases are studied. Funane et al. also state that their imaging system can be applied to other cleared organs and tissues that are in centimetre size. In similar way, Nie

(22)

et al. (2019) imaged a mouse brain in whole with multiangle-resolved subvoxel SPIM (Mars-SPIM) which combines more straightforward imaging than multiview SPIM and more complicated computational reconstruction. As mentioned above, scattering of the light appears to be a problem also when other tissue types are imaged with SPIM-based methods. Nie et al. (2019) were able to reconstruct the whole brain with better resolution and contrast than basic SPIM has achieved. They could also identify different regions of the brain and determine the density of the neurons in each region. According to Nie et al.(2019), the Mars-SPIM -system could also be applied to other organs and it has a high potential to work as a significant method when whole-brain atlases are created. In addition, it could be useful for other histological applications, such as histopathological ex- aminations in support of diagnosis.

SPIM can be modified into open-top surface microscopy which has already been applied to breast tissue imaging during a surgical operation. Glaser et al. (2017) developed a light-sheet microscope with conventional perpendicular design that images the sample from underneath at 45 degrees angle. They imaged the surface of a stained breast tissue sample that was in scale of several centimetres. The imaging time was relatively short, less than 1 minute which would be a suitable amount of time if the system was used during a surgical operation as a diagnostic method. They were able to identify both be- nign tissue areas and malignant tumour areas from the surface-images. When compared with conventional histology samples that were fixed with formalin, embedded in paraffin and stained with H&E, the quality appeared to be very similar. In addition to the breast tissue surface, they performed a volume reconstruction to a core-needle biopsy of a prostate with their light-sheet microscope. They claimed that the system would be a prac- ticable method, for example, in margin status identification of breast tissue samples as a surface microscopy method and in visualizing complicated tissue structures in three dimensions. However, the 3D reconstructed sample was very thin in volume (1 millime- tre) and had to undergo a time-consuming clearing procedure before the imaging, whereupon it cannot be directly assumed that the method could succeed in imaging samples with greater volumes. On the other hand, as the group successfully reconstructed a core needle biopsy of a prostate and it is also commonly used in breast cancer diagnosis, SPIM would be also useful in such purpose.

The usefulness of SPIM has been proven with other additional applications, such as imaging the follicles of a pig’s ovarian with higher resolution and speed than conventional ultrasound microscopy (Lin et al. 2016). In addition, the use of SPIM for studying tumour genesis and formation of metastases has been implemented in several different studies.

For instance, Lloyd-Lewis et al. (2016) used SPIM to image mammary gland epithelia

(23)

cells of a mouse with different clearing methods, and resulted in 3D reconstructions usable for further understanding of both normal and tumour tissue development of mammary gland. In similar way, Alladin et al. (2020) studied formation of a breast cancer tumour by tracking cell-cell interactions for 4 days within 3D cell culture of mammary gland cells. The metastasis formation was studied by Asokan et al. (2020) by performing SPIM imaging to track the movement and interactions of breast cancer and leukemic cells in circulation of a zebrafish embryo for 30 hours. Figure 4 presents breast cancer cells that are labelled with GFP and therefore are seen in green. On the left, the cancer cells are present as single cells, in the middle they are loosely together and on the right they have formed a cell cluster.

Figure 4. Breast cancer cells in circulation of a zebrafish embryo, imaged by SPIM.

The cancer cells appear green, as the structure of the embryo appears ma- genta. (Asokan et al. 2020)

All in all, the use of SPIM can lead to deeper understanding of cancer formation, for instance, in breast cancer research since samples of appropriate size can be imaged on the whole and tracking of cells can be performed with high resolution and continuous scanning which is a great advantage when compared with existing methods, such as MRI.

4.3 Applications of SBF-SEM

The SBF-SEM method is most usable for histological specimen in smaller size range than in micro-CT or SPIM -methods. SBF-SEM has been successfully utilized in imaging, for instance, cell populations, axons, cell ultrastructure and structures of different cell organelles in nanometre scale. The lateral resolution it can achieve is approximately 5 nanometres and the axial resolution is maximized with slice thickness of 20–25 nanometres (Peddie & Collinson 2014; Smith & Starborg 2018).

As an example of using SBF-SEM in imaging larger volumes, Svensson et al. (2017) studied the tendon structure of a mouse by cutting it into 100 nm thick slices between

(24)

electron microscopy imaging. They achieved to differentiate collagen fibrils of the tendon and study the length of them. However, the fibrils of a human tendon, which they also attempted to image, turned out to be too long of a research topic since they did not achieve desired resolution and the fibril tracking became hard due to movement of indi- vidual fibrils across the field of view.

Brain tissue and nerves are suitable research topics of SBF-SEM because they are more conductive compared to other tissue types. Hence, the nervous tissue does not require any special preparation to make the tissue more conductive in order to remove charging which moreover causes artefacts to the images. For instance, Kornfeld et al. (2017) stained three axons and SBF-SEM-reconstructed them to study the connectivity of them for better understanding of a bird’s ability to chirp which seems to be a similar process than a human’s ability to form phonemes with very precise timing. Despite that SBF-SEM was not able to find every computationally estimated synaptic connection between the cells, they still managed to discover one sixth of them. They could also study the correlation between the distance from the cell’s soma and the amount of both inhibitory and excitatory synapses based on the SBF-SEM data.

An example of using SBF-SEM to form a 3D reconstruction at cellular and ultracellular level, Pinali et al. (2013) presented the structure of cardiac myocytes of a rat and a sheep in three dimensions imaged with SBF-SEM. The attained resolution of their imaging system was 15 nm at lateral axes and the slice thickness was 50 nm which was suitable for both studying the sarcoplasmic reticulum and t-tubule structure within the myocytes. In addition, by adjusting the cutting depth and the lateral resolution to be bigger, they could also present the cellular structure of the myocardium. The second example of SBF-SEM application by Vihinen et al. (2013) demonstrated the volumetric structure of the endoplasmic reticulum (ER) of the human hepatoma cells. The ER was stained in accordance with a conventional SBF-SEM heavy-metal staining protocol which improved the resolution so that the ER could be visualized in detail. However, the ribosomes on the ER surface were not able to be visualized with SBF-SEM, while another imaging technique, electron tomography, succeeded to illustrate both the smooth ER and ribosomes. They state that the main potential of the SBF-SEM is to image many cells at a time with resolution that reaches the size of the cell organelles which enables, among other things, the determination of the amounts of different cell organelles. In addition, SBF-SEM was poorer method compared to electron tomography when membrane connections were researched. Figure 5 shows the structure of ER. On top left, there is an overall view of the cells, and on bottom right, the ER is shown in yellow (ER sheets) and lilac (ER tubules). On the right, the ER structure is presented in yellow in 3D.

(25)

Figure 5. The ER of human hepatoma cells imaged by SBF-SEM. a) An overall structure of the cells. b) The 3D structure of ER sheets in yellow. c) The 3D structure of ER sheets in yellow and ER tubules in lilac. (Vihinen et al. 2013) SBF-SEM is usually performed to soft tissue types since it makes the cutting and the controlling of conductivity more simple. However, Goggin et al. (2020) used it to visualize osteocytes in order to understand the mechanical signals between them, osteoblasts and osteoclasts in more detail because the bone development and resorption were not fully understood. In addition, the most used high-resolution method in bone imaging, micro-CT, was not able to visualize the osteocytes in ultracellular level. The group was able to reconstruct the 3D structure of a osteocyte and the distribution of them in bone matrix with higher resolution than in micro-CT which they estimated to be useful in further understanding of mechanical interactions between bone cells, for instance, in modelling the development of osteoporosis. The resolution of SBF-SEM was approximately 10 nm lat- erally and 25 nm axially, whereas micro-CT can usually reach axial and lateral resolution of 1–10 μm. Another advantage of SBF-SEM was that the obtained images were spon- taneously aligned well to each other because of the way of imaging the stationary block face instead of multiple slices separately. However, the destructiveness, complex preparation process and non-isotropic voxel data that is generated during SBF-SEM appeared to be the drawbacks compared to micro-CT in bone tissue imaging.

In addition to cellular and sub-cellular imaging achieved with SBF-SEM, it can also be applied to even smaller structures, such as chromosomes. For instance, a survey by

(26)

Chen et al. (2017) covered 3D imaging of a human cell during prophase in which the chromatin condenses into chromosomes that consist of two adjacent chromatids, or at least the studies in two dimensions suggest that. When they performed 3D reconstruction and studied the chromosomes, most of them observed the principle mentioned above.

In addition, the chromosomes had a clear 3D structure and the bigger ones resembled letters S or C and the smaller ones had a structure more similar with letter X. However, either a small amount of the chromosomes broke down during the preparation stage or the SBF-SEM system was not able to slice those chromosomes since the volume reconstruction could not be completed for them. All in all, SBF-SEM is applicable to nuclear and chromosomal research since the obtained results were compatible with current knowledge.

The SBF-SEM technology is constantly evolving and the quality of the obtained images is increased further. In addition, the scale of suitable tissue types and sizes is constantly expanded. The greatest disadvantages of the method are the destructiveness of it and the maximum sample size of approximately 1000–10 000 000 μm³ which is mostly limited by the measurement setup and the size of ultramicrotome. (Peddie & Collinson 2014) There are also other 3D electron microscopy techniques available, such as serial section transmission electron microscopy in which the tissue is sliced beforehand and the slices are imaged separately. Compared with such techniques, SBF-SEM is more easy to use since the imaging system operates automatically after adjustments and obtained images are more easily aligned during the reconstruction process since the stationary block is imaged.

(27)

5. BREAST CANCER HISTOLOGY

The following chapters provide information about histology that is needed in breast cancer diagnosis. After that, the applications of micro-CT in the field of breast cancer are studied and possibilities of SPIM and SBF-SEM are discussed based on literature presented in chapter 4.

5.1 Standard 2D histology of breast cancer

When breast cancer is diagnosed, the breast is usually examined and palpated by a doctor, after which the breast is imaged using mammography and, if necessary, with ultrasound. In addition, a large-core needle biopsy is collected from the tumour tissue.

The cancer type is classified by a histological analysis of the sample, by determining which cell type the cancer cells resemble the most. The most common breast cancer types are ductal and lobular carcinoma, the former meaning cancer that is developed in the cells of the milk ducts and the latter meaning cancer that begins in the lobules that produce milk. The invasive ductal carcinoma is the most common breast cancer since 70–80 % of invasive breast cancers are that type. (Heikkilä & Kärjä 2019)

When cancer cells are detected from a histological sample, a cancer cell is usually different than normal cell in size or in shape. Cancer cells also vary from each other within the sample. In addition to the appearance of cells, changes at the level of cell nucleus can also be noticed. A cancerous nucleus is usually bigger in size and darker than normal cell nucleus because of bigger amount of DNA, and this can be also noticed when typical staining methods are in use. As cells of the breast tissue normally form ducts and lobules, cancer cells can be noticed to form deformed structures or clusters of cells that are not organized at all. (American Cancer Society, 2015)

As the histological type of cancer has been determined, the grade of the cancer is also classified by a Bloom–Richardson system which grades the cancer based on three cri- teria: the amount of tissue that forms tubular structure, the mitotic count and the pleomorphism of the cell nuclei. Each of the categories is scored from 1 to 3 points, and the overall score determines the tumour grade from 1 to 3, the lowest indicating nearly normal tissue and the highest the most abnormal and most poorly differentiated tissue. The first criterion, the tubular structure, is usually found in normal breast tissue since the cells form ducts and therefore such tissue gets 1 point from that category, whereas cancer cells do not form that much of such organized structures, thus earning higher points. The

(28)

mitotic count is determined as the amount of dividing cells in certain field of view which appears to be higher in cancerous tissue, resulting in higher points. The last category, the nuclear pleomorphism, covers examination of cell nuclei that differ from normal nuclei in size or shape. For instance, irregular large nuclei or higher amount of nucleoli within the nuclei leads to higher points. (He et al. 2011; Heikkilä & Kärjä 2019) Figure 6 shows an example of H&E stained sample of grade 1 ductal carcinoma. The tumour can be seen as purple area on bottom right of the figure since ductal carcinoma usually appears as solid groups or cords of tumour cells. Adipose tissue (in the middle), which is also called fat tissue, forms the majority of the breast volume and stroma (top right) is a part of the tissue which has a connective role. (The human protein atlas)

Figure 6. H&E stained histology slice of ductal breast carcinoma in which the tu- mour can be seen on bottom right. (The human protein atlas)

In addition to the cancer grade, the stage of the cancer is also determined before treatment. It contains defining the tumour location and size using TNM-classification where T refers to primary tumour, N to regional lymph nodes and M to distant metastases. Sim- plified, the idea is to measure the tumour size and determine if the cancer cells have spread to regional lymph nodes or other parts of the body. To perform TNM-classification, histological analysis is needed. In addition, the histological types of breast cancer, such as ductal and lobular carcinoma, can be furthermore divided into different molecular and genetical subtypes based on gene profile analysis. However, since the gene profile analysis is not very common in clinical use, a rough estimation can be performed by immunohistochemical staining of oestrogen receptors, progesterone receptors and human epidermal growth factor 2 (HER2) which is an oncogene. Pathologist reports the

(29)

results as the proportion of positively stained nuclei, and a proper treatment method can be chosen based on the receptor status. Normal breast tissue has receptors for oestrogen and progesterone, and some breast cancer cells have also one or both of these receptors which makes hormone therapy treatment possible in which the hormone levels are lowered or drugs are used to stop a certain hormone from affecting the cancer cells.

Similarly, some breast cancer cells have too many HER2-receptors on their cell membrane which promote the growth of the cancer cells. There are various very effective targeted treatments available for HER2-positive cancers. HER2-status can be either determined by immunohistochemistry or by fluorescence in situ hybridization of the HER2- genes that may be present with too many copies. (Heikkilä & Kärjä 2019)

The earliest signs of breast cancer are small calcium deposits, calcifications, within the breast tissue. While they are very common and usually not malignant, it is important to be able to detect and distinguish suspicious calcifications from normal cases since they can also indicate invasive cancer types. They are usually noticed on a mammogram as white dots and, based on their size, calcifications are divided into macro- and microcal- cifications. Most of the detected calcifications are diagnosed as ductal carcinoma in situ (DCIS) which is the earliest form of breast cancer and not invasive. However, Mordang et al. (2018) studied the sensitivity of mammogram in calcification detection and the sensitivity was noticed to be only 45,5 % which could be improved. They stated that the sensitivity should be improved since the current stage of it leads to 68,4 % of the cancer- related calcifications to be diagnosed not until the cancer has developed into invasive stage. This could be achieved, for instance, by combining some other imaging technique with mammogram or by developing a specific method that could detect early signs of cancer with higher sensitivity while keeping the amount of false positives at current level.

Table 1 aggregates all of the histological methods that are in use when breast cancer is diagnosed.

(30)

General analysis

- variation in cell size or shape - variation in nucleus size or shape - variation in arrangement of cells

Cancer type Origin: in which cell type the cancer is developing

Cancer grade

Bloom–Richardson system - tubular structures - mitotic count

- pleomorphism of the cell nuclei

➔ Scoring from 1 to 3

Cancer stage

TNM-classification - tumour size

- regional lymph node involvement - presence of metastatic spread Molecular subtype /

receptor status

Estimation by immunohistochemistry - oestrogen receptors

- progesterone receptors - HER2

Calcifications Presence of micro- and macrocalcifications and their malignancy

Next, the possibilities of 3D histology methods for detection of breast cancer are studied and analysed.

5.2 Micro-CT in breast cancer imaging

During the recent years, micro-CT has become more popular in imaging of breast cancer samples. Research shows that micro-CT has many advantages in determining the size of the cancer and in detecting margin positive samples where the tumour tissue reaches the specimen edge. Surgeons aim to remove all of the tumour tissue when they perform mastectomy or lumpectomy in order to reduce the need for second operation, the former meaning removal of all breast tissue and the latter meaning removal of the tumour tissue.

At the same time, it is important to minimize the amount of removed normal breast tissue.

The 3D-reconstruction of the breast tissue could help the surgeon to identify and localize the tumour tissue and even help to improve the surgical technique in general.

Several piece of research show that micro-CT can locate the tumour and analyse the size of it with high accuracy. DiCorpo et al. (2020) used micro-CT to image 173 partial mastectomies. They compared the size of the tumour defined by micro-CT with results of measured gross specimen and sliced specimen. The micro-CT size and the gross

Table 1. A summary of histological study of breast cancer.

(31)

specimen size differed in average by 0.26 cm. In addition, they sliced 11 invasive type specimen after the first micro-CT-imaging and imaged the slices again both with micro- CT and optical microscope. When the analysed tumour sizes between the second micro- CT imaging and microscopy were compared, they found out that there was a high corre- spondence. The micro-CT measurements also appeared to follow a log-linear curve better than the measurements made by the pathologist. Since the log-linear curve is most likely to model the tumour size distribution, micro-CT is a potential method for tumour size detection. Moreover, Tang et al. (2015) compared cancer size measurement by micro-CT with four different methods, magnetic resonance imaging (MRI), ultrasound and mammography by imaging 50 invasive breast cancer specimens. They discovered that micro-CT had the highest accuracy of all four methods, since it had the highest correlation coefficient. However, micro-CT seemed to overestimate the cancer size in most of the cases in the same way as MRI.

One trend in research is defining the margin status of the specimen by micro-CT. Since it is crucial to remove all of the tumour, specimen edges are usually examined with a microscope after the operation in order to detect any tumour tissue touching the edges.

If cancer cells are found on the edges of the sample, it is considered as margin positive, and vice versa. In several studies, micro-CT margin status seemed to accord with the analysis by a pathologist. According to DiCorpo et al. (2020), micro-CT identified cancer touching the specimen edge in 93 % of the samples that were also considered as margin positive by the pathologist. In comparison, Qui et al. (2018) discovered 86 % correspond- ence with the pathologist statement when examining dataset of 30 lumpectomy specimen. In addition, micro-CT succeeded to detect also 47 % of the margin negative to be actually margin positive in the study by DiCorpo et al. (2020). They discovered that micro- CT identified over three times smaller area of the cancer tissue touching the specimen edge when comparing to the analysis by the pathologist, resulting in bigger amount of positive cases. This could possibly help to reduce the amount of re-operations and the risks of complications, secondary diseases and cosmetic damage. An example of a margin negative specimen, in which the tumour tissue does not reach the edges, is shown in figure 7. Nevertheless, as the margin status was studied by Qiu et al. (2018), micro- CT resulted in four false-negatives of 29 samples which was due to the young age of the patients (less than 50 years) and hence the higher density of their breast tissue. Tumours are detected as density differences which result as greyscale differences in the radio- graph. Hence, micro-CT might not be the most suitable method for young patients. This could be prevented by using a contrast agent or by performing mammography imaging before the operation in order to select patients that could benefit from micro-CT during

3D Histology Imaging and Image Reconstruction Methods

Jonna Jokela