Evaluating Advanced Machine Learning Models for Histopathological Diagnosis of Hansen Disease

Area of Science:

• Medical diagnostics
• Computational pathology
• Infectious disease research

Background:

• Leprosy, caused by Mycobacterium leprae and Mycobacterium lepromatosis, is a persistent public health issue in tropical areas.
• Machine learning (ML) presents innovative strategies for enhancing the diagnosis of this complex infectious disease.

Purpose of the Study:

• To evaluate the effectiveness of various machine learning models for the histopathological diagnosis of Hansen disease.
• To validate the utility of ML in identifying leprosy from tissue slide images.

Main Methods:

• An observational study analyzed 55 leprosy and 51 control H&E-stained skin tissue slides.
• Microphotographs were processed using Cross-Industry Standard Process for Data Mining (CRISP-DM) for ML model development.
• Five ML models were assessed, focusing on data normalization and performance metrics like accuracy, sensitivity, and specificity.

Main Results:

• The artificial neural network (ANN) model achieved 70% accuracy, 74% sensitivity, and 65% specificity.
• The ANN model's receiver operating characteristic curve showed an area under the curve of 0.71.
• Other models like decision trees and random forests had similar accuracy but lower sensitivity, increasing the risk of false negatives.

Conclusions:

• The artificial neural network (ANN) model shows potential as a tool for leprosy detection.
• Further research is needed to improve model adaptability for diverse clinical settings and patient care integration.