AI-enhanced PathCAM microscopy for rapid low-magnification detection of Tuberculosis in sputum swabs

We present PathCAM, a real-time, computer-assisted pathology microscope that enables AI-assisted detection of acid-fast bacteria (AFB) in tuberculosis (TB) diagnostics. PathCAM combines high-resolution image capture, GPU-accelerated multi-resolution compositing, and on-the-fly AI inference to support slide navigation for bacteria counting in stained sputum samples. This workflow addresses the tedious and time-intensive process of manually identifying 2–4 µm TB bacteria across dozens of high-power fields. We demonstrate AI-in-the-loop classification of M. Tuberculosis in clinical specimens across multiple magnifications. Our results suggest PathCAM can improve diagnostic efficiency while pushing the boundaries of conventional AI training paradigms in digital pathology.