WSI AFB and AI deliver highest sensitivity for TB detection

Area of Science:

• Medical Diagnostics
• Infectious Diseases
• Digital Pathology

Background:

• Tuberculosis (TB) is a chronic respiratory infectious disease caused by Mycobacterium tuberculosis (Mtb).
• Current diagnostic methods fail to detect 24% of TB patients, highlighting limitations in existing methodologies.

Purpose of the Study:

• To develop and evaluate a novel Whole Slide Imaging (WSI) platform for enhanced tuberculosis (TB) detection.
• To assess the diagnostic performance of the WSI-TB platform against conventional methods.

Main Methods:

• Development of a WSI platform with a Curved Surface Focus Algorithm (CSFA) for high-speed digitization.
• Implementation of a two-stage deep learning AI pipeline (YOLOv5 and ResNet-18) for automated acid-fast bacilli (AFB) identification.
• Prospective and retrospective evaluation of the WSI-TB platform in 1097 patients, comparing it with smear microscopy, culture, and Xpert MTB/RIF.

Main Results:

• The WSI-TB platform demonstrated an overall sensitivity of 42.43% and 100.00% specificity.
• WSI-TB sensitivity surpassed traditional acid-fast staining smear (18.80%) and culture (30.36%).
• Compared to manual microscopy, WSI-TB showed higher positive rates (42.43% vs. 18.8%); compared to culture (43.46% vs. 30.36%); and compared to Xpert (62.95% vs. 44.26%).

Conclusions:

• The WSI-TB technology significantly enhances the sensitivity of tuberculosis sputum smear testing.
• The platform maintains 100% specificity, offering a novel approach to improve TB detection rates.
• This advancement provides a new tool to address the shortcomings in current TB diagnostic methodologies.