Artificial Intelligence via Competitive Learning and Image Analysis for Endometrial Malignancies: Discriminating Endometrial Cells and Lesions

Artificial Intelligence via Competitive Learning and Image Analysis for Endometrial Malignancies: Discriminating Endometrial Cells and Lesions

Abraham Pouliakis (2nd Department of Pathology, National and Kapodistrian University of Athens, Athens, Greece), Niki Margari (Independent Researcher, Greece), Effrosyni Karakitsou (Department of Biology, University of Barcelona, Barcelona, Spain), George Valasoulis (Department of Obstetrics and Gynaecology, IASO Thessaly Hospital, Larisa, Greece), Nektarios Koufopoulos (2nd Department of Pathology, National and Kapodistrian University of Athens, Greece), Nikolaos Koureas (2nd Department of Gynecology, St. Savas Hospital, Athens, Greece), Evangelia Alamanou (Department of Obstetrics and Gynecology, Tzaneio Hospital, Piraeus, Greece), Vassileios Pergialiotis (3rd Department of Obstetrics and Gynaecology, National and Kapodistrian University of Athens, Athens, Greece), Vasileia Damaskou (2nd Department of Pathology, National and Kapodistrian University of Athens, Athens, Greece) and Ioannis G. Panayiotides (2nd Department of Pathology, National and Kapodistrian University of Athens, Athens, Greece)
Copyright: © 2019 |Pages: 17
DOI: 10.4018/IJRQEH.2019100102

Abstract

Objective of this study is to investigate the potential of an artificial intelligence (AI) technique, based on competitive learning, for the discrimination of benign from malignant endometrial nuclei and lesions. For this purpose, 416 liquid-based cytological smears with histological confirmation were collected, each smear corresponded to one patient. From each smear was extracted nuclear morphometric features by the application of an image analysis system. Subsequently nuclei measurement from 50% of the cases were used to train the AI system to classify each individual nucleus as benign or malignant. The remaining measurement, from the unused 50% of the cases, were used for AI system performance evaluation. Based on the results of nucleus classification the patients were discriminated as having benign or malignant disease by a secondary subsystem specifically trained for this purpose. Based on the results it was conclude that AI based computerized systems have the potential for the classification of both endometrial nuclei and lesions.
Article Preview
Top

Introduction

The endometrium is a tissue system that is affected by the menstrual cycle under the estrogen effects and undergoes a cyclic regeneration. Various factors such as estrogens, sex steroids, oncogenes, growth factors and numerous peptides (Murphy, Murphy, & Friesen, 1987; Shyamala & Ferenczy, 1981) can result in premalignant and malignant transformation of the endometrium. Until now the standard procedure to identify potential malignancies of the endometrium is the examination of endometrial samples via the cytomorphological evaluation of Papanicolaou stained direct smears and histological examination of dilation and curettage biological samples. The second method involves patient anaesthesia and can be painful while the cytological examination is less invasive and therefore friendlier to the patient. Especially the application of liquid-based cytology (LBC) that has largely replaced the conventional direct smear preparations the last decade has diagnostic advantages: due to the monolayer structure of the microscopy glass slides while providing the opportunity of cytomorphological discrimination of nuclear features, based on geometric and densitometric characteristics.

Machine Learning and in general Artificial Intelligence (AI) techniques have already been applied in medicine (Adams, Bello, & Dumancas, 2015; Almeida & Noble, 2000; Cochand-Priollet et al., 2006; Darcy, Louie, & Roberts, 2016; Foran, Chen, & Yang, 2011; Gaiser et al., 2010; Giacomini, Ruggiero, Calegari, & Bertone, 2000; Grabe et al., 2010; Hajmeer, Basheer, & Najjar, 1997; Karakitsos et al., 1997; Karakitsos et al., 1996; Karar & El-Brawany, 2011; Luo et al., 2015; Makris et al., 2017; Mobasser & Hashtrudi-Zaad, 2012; Mouwen, Capita, Alonso-Calleja, Prieto-Gomez, & Prieto, 2006; Pergialiotis et al., 2018; Perroti et al., 2018; Pitoglou, 2018; Pouliakis et al., 2016; Pouliakis et al., 2018; Salamalekis et al., 2019; Seffens, Evans, Minority Health, & Taylor, 2015; Siristatidis, Chrelias, Pouliakis, Katsimanis, & Kassanos, 2010; Siristatidis, Pouliakis, Chrelias, & Kassanos, 2011; Siristatidis et al., 2016; Su, Xu, He, & Song, 2016; Vilhena et al., 2017). There are numerous applications which they involve a) classical statistical models b) more advanced techniques, such as neural networks or c) fuzzy systems. In the field of medical oncology, there are numerous applications (Cruz & Wishart, 2006; Hassan, Ruusuvuori, Latonen, & Huttunen, 2015; Kourou, Exarchos, Exarchos, Karamouzis, & Fotiadis, 2015; Pouliakis et al., 2016; Simões et al., 2014). In contrast, publications based on endometrial cytological material evaluation by Machine Learning Applications (MLAs) is rather poor.

Complete Article List

Search this Journal:
Reset
Open Access Articles
Volume 9: 4 Issues (2020): 2 Released, 2 Forthcoming
Volume 8: 4 Issues (2019)
Volume 7: 4 Issues (2018)
Volume 6: 4 Issues (2017)
Volume 5: 4 Issues (2016)
Volume 4: 4 Issues (2015)
Volume 3: 4 Issues (2014)
Volume 2: 4 Issues (2013)
Volume 1: 4 Issues (2012)
View Complete Journal Contents Listing