MedScan Mentor

Welcome to MedScan Mentor: Your portal to mastering neuroimaging diagnostics. Begin by uploading a CT scan to start the analysis.
Manual Annotation Tool: Users can manually draw bounding boxes to predict tumor locations, enhancing learning through hands-on experience.
AI-driven Insight: MedScan instantly compares user annotations with AI predictions, providing immediate feedback on detection accuracy.
After the AI evaluation, users receive detailed feedback on their diagnostic accuracy, encouraging precision in medical interpretations.
Educational Summaries: Each session is complemented by a detailed report drawing on recent medical research, fostering continuous learning.
Review and Continuous Learning Interface
Review Past Cases: Users can access and review their previous analyses, promoting long-term learning and retention of diagnostic skills.

Inspiration

The inspiration behind MedScan sprang from the need to bridge the gap between theoretical knowledge and practical skills in medical education, particularly in the challenging field of neurology. Recognizing that medical students often struggle with accurately diagnosing brain disorders from imaging studies due to limited hands-on experience, we aimed to create a platform that not only assists in the detection and analysis of brain diseases through CT scans but also educates and guides them through the process.

What it does

MedScan serves as an educational tool that facilitates the learning and diagnostic skills of medical students and professionals. Users upload brain CT scans, and the platform enables them to manually annotate suspected areas of pathology. MedScan then compares these user-generated annotations with its AI model’s detections, providing immediate feedback and an educational report that details the characteristics and nuances of detected conditions. This report integrates the latest scholarly articles and data from PubMed to ensure users receive current and relevant medical information. Moreover, users can easily access their previous analysis along with generated report as future reference in their studies.

How we built it

We built MedScan using a robust stack of technologies to ensure a seamless and educational experience. The core of MedScan is powered by YOLOv5 for object detection, trained on a comprehensive dataset of MRI images annotated for various brain tumors. The backend is built on Python and utilizes MongoDB for storing user data and interaction histories, ensuring that users can track their progress and review past cases. The educational reports are generated using RAG (Retrieval-Augmented Generation) technology to pull in the latest research articles from PubMed, providing users with up-to-date medical information.

Challenges we ran into

One of the primary challenges was ensuring the accuracy and reliability of the AI model in detecting subtle and diverse brain tumors across different MRI angles. Balancing the model’s sensitivity and specificity to maximize educational value was intricate. Additionally, integrating real-time data retrieval from PubMed while maintaining fast response times for a smooth user interface presented technical hurdles.

Accomplishments that we're proud of

We are immensely proud of developing a functioning AI that not only detects brain tumors with high accuracy but also helps medical students understand their diagnostic reasoning errors through real-time feedback. Creating an interface that allows for intuitive interaction with complex medical imaging and integrating cutting-edge AI to support medical education are accomplishments that resonate with our team’s goals.

What we learned

Throughout the development of MedScan, our team gained deeper insights into the challenges of medical imaging analysis, especially in the context of educational tools. We learned advanced techniques in machine learning model training, particularly in medical image processing, and honed our skills in developing user-centric software that addresses real-world problems.

What's next for Untitled

Looking ahead, we plan to expand MedScan’s capabilities to include more types of medical imaging studies, such as PET scans and ultrasounds. We also aim to incorporate more interactive elements, such as virtual reality (VR), to create an immersive learning environment. Additionally, we are looking into partnerships with medical schools and institutions to integrate MedScan into their curricula, enhancing the practical training of future medical professionals.