Auto Machine Learning and Convolutional Neural Network in Diabetes Mellitus Research-The Role of Histopathological...

Area of Science:

• Reproductive biology
• Computational pathology
• Diabetology

Background:

• Histopathological images are crucial for diagnosing complex pathologies.
• Existing computational models often depend on limited public datasets.
• Diabetes mellitus (DM) can significantly impact female reproductive health.

Purpose of the Study:

• To create an original histopathological image dataset from a female rat model of diabetes mellitus.
• To evaluate the effects of an antidiabetic synthetic compound (AD_SC) on reproductive organs (vagina, uterus, ovary).
• To develop and apply AI models for classifying histopathological images and extracted features.

Main Methods:

• Acquisition of histopathological images from normal, DM-induced, and AD_SC-treated female rats.
• Development of a custom-built convolutional neural network (CB-CNN) for image classification.
• Extraction and classification of textural features (contrast, entropy, energy, homogeneity) using PyCaret Auto Machine Learning (AutoML).

Main Results:

• Uterine tissue classification achieved 94.5% accuracy for DM and 85.8% for AD_SC groups.
• Linear Discriminant Analysis (LDA) using vaginal tissue features yielded 86.3% accuracy.
• The study demonstrates the potential of AI in analyzing reproductive tissue changes in diabetes.

Conclusions:

• AI algorithms, including CNN and machine learning, are effective for classifying histopathological images of reproductive tissues.
• The developed dataset and methods provide valuable insights into the impact of diabetes and its treatment on female reproductive health.
• This research highlights the utility of computational approaches in advancing reproductive pathology research.