Automated Microscope for TB Detection

The Challenge

Manual TB sputum smear microscopy is slow and error-prone, requiring trained lab technicians to examine slides for hours. Misdiagnosis rates in developing countries can exceed 20%.

Approach & Solution

Designed a custom automated microscope system that captures high-resolution sputum smear images and processes them through a deep CNN pipeline. Built the imaging hardware integration, trained models on thousands of labeled microscopy images, and optimized inference for real-time detection.

Results & Impact

• Achieved 95%+ detection accuracy on test datasets
• Reduced slide analysis time from 30+ minutes to under 2 minutes
• Published as Master's thesis at Air University
• System capable of processing 50+ slides per hour vs 3-4 manually