Computer aided detection of tuberculosis using two classifiers.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer Science

Background:

• Tuberculosis (TB) diagnosis is challenging in underdeveloped countries due to limited tools and high case numbers.
• Current methods like microscopy and X-rays are tedious and prone to miss-diagnosis.
• There is a need for automated, accurate TB detection systems.

Purpose of the Study:

• To develop and evaluate an Artificial Intelligence (AI)-driven Computer Aided Detection (CAD) system for tuberculosis detection.
• To address the limitations of manual diagnosis by employing deep learning models.
• To improve the accuracy and efficiency of tuberculosis identification from medical images.

Main Methods:

• Utilized pretrained AlexNet models for automated discrimination of tuberculosis cases.
• Employed both chest X-ray datasets (Kaggle) and microscopic slide images (hospital and Kaggle).
• Classified images using AlexNet combined with Softmax and Support Vector Machine (SVM) classifiers.

Main Results:

• AlexNet+SVM achieved 98.73% accuracy for microscopic slide classification.
• AlexNet+Softmax achieved 98.19% accuracy for chest X-ray classification.
• The AI models demonstrated superior performance compared to existing studies.

Conclusions:

• AI-driven CAD systems show significant promise for accurate and efficient tuberculosis detection.
• The developed models outperform current diagnostic benchmarks.
• Future work includes integrating Internet of Medical Things (IoMT) for an enhanced diagnostic platform.