Fluoreszcens in situ hibridizációval igazolt onkopatológiai génelváltozások automatizált kiértékelése digitalizált teljes mintákon = Automated evaluation of oncopathological genetic aberrations in whole slide images of digitalized samples detected by using fluorescence in situ hybridization

Csizmadia, Annamária and Molnár, Béla and Kapczár, Dóra and Papp, Gergő and Krenács, Tibor (2024) Fluoreszcens in situ hibridizációval igazolt onkopatológiai génelváltozások automatizált kiértékelése digitalizált teljes mintákon = Automated evaluation of oncopathological genetic aberrations in whole slide images of digitalized samples detected by using fluorescence in situ hybridization. SCIENTIA ET SECURITAS, 5 (2). pp. 146-158. ISSN 2732-2688

Preview

Text 112-article-p146.pdf - Published Version Available under License Creative Commons Attribution. Download (3MB) | Preview

Abstract

Fluoreszcens in situ hibridizációval (FISH) genetikai aberrációk igazolhatók tumormintákon. Digitális mikroszkópiára épülő automatizált képanalízissel a nagy felbontású FISH génjelek mintánként több ezer sejtben igazolhatók, és a kezelést befolyásoló daganatheterogenitás is pontosan meghatározható. A módszer a patológus munkaterhelésének csökkentése mellett támogatja a hatékonyabb diagnózist és az erre épülő onkológiai terápiát. A FISHQuant algoritmus finombeállítás után nagyszámú tumormintán igazolta a módszer alkalmasságát diagnosztikus célokra, mind génátrendeződéses, mind kópiaszám-eltéréses génhibák megbízható kimutatásában. Vizsgálataink ugyancsak rávilágítottak a 3D magszegmentálás előnyeire a 2D módszerrel szemben. A dolgozatban röviden bemutatjuk kutatásunk néhány eredményét. | The fluorescence in situ hybridization (FISH) technique has been widely applied in molecular pathology for the in situ identification of well-known genetic aberrations in formalin-fixed paraffin embedded tissues, frozen sections, and cytological smears. In tumor pathology, FISH provides qualitative and quantitative results to support differential diagnosis often with a direct therapeutic consequence. It can be used to identify the genetically affected cells and their proportion within tumor populations. FISH can determine genetic abnormalities at a single cell level within the morphological complexity of a tumor. Advanced digital microscopy combined with image analysis algorithm offers a great chance to extract all clinically useful data from whole FISH samples. We have been developing a software tool called FISHQuant to support pathology diagnostics by semiatomated digital image analysis of major genetic abnormalities characterized by numerical and steric deviations. Based on HER2 immunohistochemical positivity breast cancer samples were selected for FISH. A dual color HER2/Cep17 probe was used, which can detect HER2 gene copies in relation to chromosome 17 to idenfify tumor cells with increased (>2 copy/chromosome) ratios of HER2 signals. In addition to the specific genetic aberration examined, tumor cell populations also carry non-specific deviations, which can be accurately detected with automated digital analysis, but might remain hidden with the traditional microscopy. Although our image analysis found similar HER2 copy number abnormalities to the eye controltesting, the results gained in tissue sections should be treated with caution due to the random truncation related signal loss introduced by cutting. Cultured breast cancer cells and their cell block sections offered chance for reliable nuclear segmentation and to compare FISH analysis in tumor cells in intact and cut through cell nuclei. We used SK-BR-3 a HER2 overexpressing cell line, MDA-MB-453 a triple-negative, and ZR-75-1 a HER2 2+ positive cell line. FISHQuant can also be used for determining sterical genetic abnormalities, i.e. gene translocations/fusions. For the diagnostic interpretation of the small signals their exact number and relative position within the nucleus must be determined accurately. At present our FISHQuant algorithms use 2D images for the nuclear segmentation, which, in high cell density tissue sections may results in suboptimal separation of cell nuclei. For more accurate rendering of FISH signals to individual cells we have been testing 3D nuclear segmentation algorithms comparted to our 2D based method in ~5 µm-thick samples. The potential of Cellpose, an artificial intelligence based image anatomical open source 3D segmentation and learning algorithm designed and trained on wide range of image datasets, have been analyzed. Our studies comparing different nuclear segmentation tools highlighted the advantages of 3D nucleus segmentation compared to the 2D algorithm.

Actions (login required)

Edit Item