A Novel Histological Dataset and Machine Learning Applications

Abstract

Histology has significant importance in the medical field and healthcare services in terms of microbiological studies. Automatic analysis of tissues and organs based on histological images is an open problem due to the shortcomings of necessary tools. Moreover, the accurate identification and analysis of tissues that is a combination of cells are essential to understanding the mechanisms of diseases and to making a diagnosis. The effective performance of machine learning (ML) and deep learning (DL) methods has provided the solution to several state-of-the-art medical problems. In this study, a novel histological dataset was created using the preparations prepared both for students in laboratory courses and obtained by ourselves in the Department of Histology and Embryology. The created dataset consists of blood, connective, epithelial, muscle, and nervous tissue. Blood, connective, epithelial, muscle, and nervous tissue preparations were obtained from human tissues or tissues from various human-like mammals at different times. Various ML techniques have been tested to provide a comprehensive analysis of performance in classification. In experimental studies, AdaBoost (AB), Artificial Neural Networks (ANN), Decision Tree (DT), Logistic Regression (LR), Naive Bayes (NB), Random Forest (RF), and Support Vector Machines (SVM) have been analyzed. The proposed artificial intelligence (AI) framework is useful as educational material for undergraduate and graduate students in medical faculties and health sciences, especially during pandemic and distance education periods. In addition, it can also be utilized as a computer-aided medical decision support system for medical experts to minimize spent-time and job performance losses.