Comprehensive Computer-Aided Decision Support Framework to Diagnose Tuberculosis From Chest X-Ray Images: Data Mining...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Tuberculosis (TB) is a fatal infectious disease requiring early diagnosis.
• Computer-aided diagnosis (CAD) tools for TB from chest radiographs (CXRs) exist but lack descriptive information.
• Existing methods offer binary classification (TB positive/negative) without detailed insights.

Purpose of the Study:

• To develop a comprehensive CAD framework for effective TB diagnosis.
• To provide visual and descriptive information from a patient database.
• To enhance diagnostic decision-making for radiologists.

Main Methods:

• A fusion-based deep classification network was proposed for CAD decisions.
• A multilevel similarity measure algorithm using multiscale information fusion was devised.
• The framework retrieves best-matched cases from a previous patient database.

Main Results:

• The framework was evaluated on two prominent CXR datasets.
• The classification model achieved high performance metrics (e.g., 0.965 AUC).
• The proposed system outperformed various state-of-the-art methods.

Conclusions:

• A comprehensive CAD framework for TB diagnosis from CXRs was presented.
• The framework retrieves relevant cases and clinical observations, aiding radiologists.
• Retrieval results support radiologists in making effective diagnostic decisions and subjective validation.