Xpert merges deep-learning diagnostics with immersive AR visualization and adaptive learning to transform how medical students understand pneumonia.
View Our Journey →A dual-engine telemedicine platform that combines CNN-powered pneumonia detection with an intelligent educational layer and AR visualization.
Current pneumonia detection tools offer static 2D predictions with no educational context. Medical students lack immersive tools to understand pathology progression and clinical reasoning.
Xpert combines a dual AI pipeline (U-Net segmentation + DenseNet classification) with AR-powered 3D visualization and an adaptive learning engine. The platform doesn't just detect pneumonia — it teaches, explains, and adapts to each student's competency level through Grad-CAM heatmaps, a RAG-powered chatbot, and the MAPE-K feedback framework.
Medical students learning through AI-guided simulation, physicians using AR visualization for deeper case analysis, and general practitioners needing quick diagnostic reference overlays.
Six integrated modules that power the Xpert experience from upload to insight.
CLAHE enhancement, U-Net segmentation, and DenseNet121 classification with CheXNet weights. End-to-end inference under 5 seconds.
Gradient-weighted class activation maps that highlight clinically relevant regions on the X-ray for transparent, explainable diagnostics.
Unity-powered AR rendering of lung anatomy with heatmap overlays, bounding boxes, and interactive 3D manipulation via WebGL.
ChromaDB vector storage with Gemini 2.5 Flash for grounded, citation-backed medical responses with built-in hallucination prevention.
MAPE-K framework that tracks student performance, identifies competency gaps, and generates personalized clinical reports.
Student vs. Admin role separation, token-based authentication, PII-safe logging, and OWASP Top 10 compliance.
Interface screenshots from the working prototype.
Five developers, designers, and engineers from the University of Wollongong in Dubai, building the future of medical education.
Builds the student and doctor web portals, role-based UI, file upload flows, and real-time processing indicators. Enforces platform security through RBAC, PII protection, and OWASP-compliant hardening.
Leads technical enforcement, repository administration, AI pipeline validation, RAG integrity testing, and Grad-CAM heatmap verification.
Manages database architecture, medical dataset pipelines, ChromaDB vector storage, and data integrity across the full inference workflow.
Develops the Unity AR visualization system, converting AI outputs into interactive 3D lung models with heatmap overlays and WebGL rendering.
Leads user research, validation surveys, heuristic usability evaluations, and Xpert's visual branding and demo presentation strategy.
Dr. Milan Dordevic
University of Wollongong in Dubai
The Xpert team with Dr. Milan Dordevic
From feasibility study to final demo — every step of our journey building Xpert.
Migrating from SQLite to MySQL with Azure Blob/AWS S3 for DICOM storage, and deploying the FastAPI backend to Vercel or Railway for 24/7 access across medical institutions.
Expanding beyond pneumonia to detect cardiomegaly, pleural effusion, and tuberculosis through an ensemble of specialized vision models.
Automating ingestion of real-time medical journals and textbooks into the ChromaDB vector database to keep the educational assistant aligned with the latest clinical guidelines.