Within Top 25 of Hacknosis Hackathon
MediFlow is a cutting-edge EHR system designed to empower healthcare providers with powerful data analytics and visual presentations. MediFlow strives to design an efficient and streamlined workflow for physicians in order to reduce patient wait time, increase diagnosis accuracy, and overall enhance patient experience.
Our platform leverages advanced techniques in NLP, medical image preprocessing, and image segmentation to extract valuable insights from unstructured healthcare data, including clinical notes, diagnostic imaging reports, and lab results.
Using AI to transform the Medical field
The project is divided into 3 Microservices, namely frontend, backend, and ai-engine, and maintained separately on GitHub.
For frontend, we chose Angular material UI to build a modularized web portal for doctors, for web server we picked Spring Boot, making use of Spring Security’s features for a secured Role-based HTTP communication between frontend and backend. Spring JPA is used along with MySQL database for an abstracted data access and update. Spring Security is used to impose a riguous Role-Based Access Scheme, ensuring user privacy and system security.
Moreover, OpenText’s Content Storage Service, Publication Service, Viewing & Transformation Service are core aspects of our implementation. The spring web server communicates with the Content Storage Service through OAuth2 sessions, during which the newly uploaded report is sent to CSS for virus scan and long-term storage. The Publication Service and Viewing Service is used togethe r to provide a seamless integration of Brava! image viewer into our Angular web app.
Furthermore, AWS Medical Comprehend is used to generate textual insight of provided text content, the Spring web server connects with AWS services via a secured web client.
Finally, for AI engine, we utilized Django, TorchIO, and PyTorch to provided image deblurring and denoising functionality, making the medical images clear to see by doctors.