AI-poweredskin carein the palm of your hand
Completing the Puzzle for the Largest-ever Dataset on Skin NTDs
Setting the stage for high-quality, multimodal, data-driven insights
SkincAir is conducting a large-scale data collection exercise with the aim of developing an innovative, AI-driven mobile application for detecting and predicting skin Neglected Tropical Diseases (NTDs).
The resulting dataset will encompass tabular data (e.g. patient demographics, medical history, and treatment details), images, and geospatial inputs. A harmonised electronic Case Report Form (eCRF) will ensure uniformity in the type and format of the data collected for the validation study and piloting on the ground.
The data goal
3,500
High-resolution images of skin NTDs with depth and scaling information, clinical and epidemiological data
12
Hospitals involved for data labelling, annotation, diagnosis check and image segmentation
Our metadata dictionary
AI Models for Early Detection, Prognosis & Treatment of Skin NTDs
Building, fine-tuning and validating the SkincAIr detection model
AI-powered diagnostic support
SkincAIr is developing an advanced AI model for detecting skin conditions using medical images. This technology will use deep learning to analyse images of the skin captured via smartphone and provide accurate diagnostic support to frontline healthcare workers, particularly in low- and middle-income countries where medical resources are limited.
Fair and inclusive model
Our goal is to improve the early detection of skin NTDs, inform treatment decisions, and ultimately reduce their impact by enabling faster and more reliable diagnoses. The model will be trained using diverse datasets collected from multiple regions to ensure an optimal performance across different skin tones, age groups.
Smart & scalable tech
To ensure fairness and accuracy, we are incorporating bias detection methods and following a standardised image collection protocol. Built on a robust data pipeline, SkincAIr uses state-of-the-art AI architectures, such as convolutional neural networks and vision transformers, while also integrating real-world patient data to enhance diagnostic accuracy.
Data Collection
Quality Control
Data Preparation & Processing
Addressing shortcut learning & bias
Model Development
Integration of meta data
Model Optimization
Model Evaluation
Final Models for Detection & Prediction of Skin NTDs
Output
A Reliable, Flexible Mobile App for Frontline Health Workers
The power to detect, monitor and manage skin NTDs from your own device
The SkincAIr mobile app will support frontline health workers in the early detection and tracking of skin NTDs using their smartphones. Their concerns and first-hand input will drive the co-design process for the app, and their feedback will continuously shape how it evolves and adapts to patients’ needs in a real-world setting. The app will run on Android, iOS, and web platforms, based on a modular architecture designed for high reliability, scalability, maintainability, and fast performance.
Key features
Real-time image capture
Augmented vision with built-in guidance to help frontline health workers take highquality skin photos, even in remote or lowconnectivity settings.
Augmented detection
Captured images are then processed directly on the mobile device, compressed, and fed into our AI model for near-instant diagnostic feedback.
QR code generation for easy access
To support follow-up treatment, the app will generate QR codes in positive diagnosis cases. This will enable hospitals to instantly access patient data and offer benefits such as treatment discounts during clinical validation.
Dedicated training resources & capacity building
In addition to its diagnostic capabilities, the app features an educational section offering multilingual training materials and quizzes to help FHWs enhance their knowledge and skills. It also collects geospatial data, which is visualised on public health dashboards to map disease outbreaks and inform health interventions.