Empowering a sub-Saharan telemedicine leader with 50,000+ human-validated medical images to detect tropical diseases with 28% higher precision.

Client Overview

A leading telemedicine provider in sub-Saharan Africa sought to develop an AI-powered diagnostic tool for early detection of tropical diseases using mobile-captured images. Facing challenges with inaccurate global datasets that overlooked local skin tones and environmental factors, the client partnered with DataLens Africa for specialized data annotation services.

Project Scope

DataLens handled the annotation of 50,000+ medical images, including X-rays, ultrasounds, and smartphone photos of skin lesions. Services included:

• •Object detection for anomalies
• •Semantic segmentation for tissue boundaries
• •Multi-label classification for disease indicators

Annotations incorporated African demographic diversity, with human validators ensuring cultural and contextual relevance.

Challenges

• •Variability in image quality from low-resource settings.
• •Bias in pre-existing datasets, leading to false positives/negatives for African patients.
• •Need for rapid turnaround to support field deployments.

DataLens Solution

Using the DataLens AI Studio, a team of 20 domain-expert annotators (including certified radiologists) applied multi-layer quality checks. Custom guidelines addressed nuances like lighting variations in rural clinics. The process integrated human-in-the-loop validation with automated pre-labeling for efficiency.

Results

• •Achieved 95% annotation accuracy, validated against expert reviews.
• •Enabled the client's AI model to improve diagnostic precision by 28% on African test sets.
• •Project completed in 8 weeks, reducing model training time by 40% and supporting deployment to 500+ clinics.