Breast cancer is one of the most prevalent and life-threatening diseases affecting millions of women worldwide. Early detection plays a crucial role in improving survival rates and treatment outcomes.
This demo video provides a comprehensive walkthrough of our AI-powered breast cancer detection system, demonstrating how artificial intelligence can assist in early detection and accurate diagnosis. The platform is designed to be user-friendly and efficient, ensuring accessibility for patients and healthcare professionals alike.
The demo begins with an introduction to our website’s homepage, showcasing its clean design, intuitive navigation, and core functionalities. Users are guided through the primary sections, including:
✅ "Start Check Process" Button – Directs users to the mammogram upload section.
✅ Navigation Bar – Easy access to educational resources, doctor consultation, and support.
This section demonstrates how users can upload their mammograms for AI analysis. The steps include:
🔸 Selecting and uploading an image of the mammogram.
🔸 AI processing the image using our custom-trained model with 15GB of real patient data.
🔸 Generating results based on AI’s detection, classifying the scan as:
High Risk (Positive) – Requires immediate medical attention.
Moderate Risk – Suggests further clinical evaluation.
Low Risk (Negative) – No immediate signs of abnormality detected.
Users are assured that the AI model is manually trained for accuracy, offering 73% precision, significantly outperforming datasets like SurfaceNet and GitHub, which show only 26-30% accuracy per case.
Once the AI has analyzed the uploaded mammogram, the system redirects users to a detailed result page, which includes:
✅ Clear categorization (Positive, Moderate, or Negative) with color-coded visual indicators.
✅ Explanations and recommended next steps based on the AI’s prediction.
✅ User-friendly layout to ensure easy understanding of medical insights.
The AI-driven approach enhances accuracy and reduces the chances of misinterpretation compared to traditional screening methods.
The demo then explores the Educational Resources section, where users can access:
📌 Breast Cancer Awareness Guides – Symptoms, risk factors, and early warning signs.
📌 Prevention Tips – Healthy lifestyle habits and regular screening importance.
📌 Self-Examination Instructions – Step-by-step guidance on self-screening techniques.
By providing these resources, our platform ensures that users are well-informed and proactive about their health.
A critical feature of our platform is the ability to consult a doctor directly through the website. The demo showcases:
👩⚕️ How users can navigate to the "Consult a Doctor" page.
📅 Booking an appointment for expert medical advice.
📩 Connecting with certified professionals for further evaluation and guidance.
This feature ensures that users receive timely medical support based on their test results.
For users who need assistance, the demo highlights the "How to Use" section, explaining:
🔹 Step-by-step instructions on uploading and interpreting results.
🔹 Frequently Asked Questions (FAQs) for troubleshooting.
🔹 Contact & Support section for technical help and medical inquiries.
This ensures that even non-tech-savvy users can effortlessly navigate the platform.
The demo video concludes by reinforcing the importance of early detection and how our AI-driven solution bridges the gap between technology and healthcare. With higher accuracy, real-time analysis, and expert consultations, this project represents a significant leap in accessible breast cancer detection.
🎥 Watch the demo now to see how AI is transforming the future of healthcare! 🚀
Bugs We Faced & How We Fixed Them in Our AI Breast Cancer Detection Project During the development of our AI-powered Early Breast Cancer Detection System, we encountered several technical challenges and bugs that we had to fix to ensure smooth functionality. Here are some of the key issues we faced and how we resolved them: 1️⃣ Image Upload & Processing Failure ❌ Bug: Initially, the mammogram image upload feature wasn’t working correctly, leading to images not being processed by the AI model. This happened due to incorrect file type validation and API misconfigurations. ✔️ Fix: We modified the backend to properly validate image formats (JPG, PNG, DICOM) and optimized API requests to ensure seamless image transmission to the AI model for analysis. 2️⃣ AI Model Prediction Delay ❌ Bug: The AI model took too long (10-15 seconds) to generate results, leading to a poor user experience. ✔️ Fix: We optimized the model by reducing unnecessary computations, implementing model quantization, and using FastAPI for real-time inference, reducing prediction time to under 5 seconds. 3️⃣ Incorrect Risk Classification Display ❌ Bug: Sometimes, the detection results showed the wrong risk level (e.g., a high-risk scan being classified as "Normal"). ✔️ Fix: The issue was traced to a misalignment in threshold values. We refined the classification logic to accurately differentiate between "Normal," "Moderate Risk," and "High Risk" based on probability scores from the model. 4️⃣ JavaScript Button Click Not Redirecting to the "Consult Doctor" Page ❌ Bug: The "Consult Doctor" button was unresponsive and didn’t redirect users to the consultation page. ✔️ Fix: The issue was due to a missing href link and incorrect event listener handling. We fixed it by correctly linking the button to "Consult a Doctor.html" and ensuring proper event binding. 5️⃣ Mobile Responsiveness Issues ❌ Bug: The website didn’t display properly on mobile devices, causing UI elements to overlap or become unclickable. ✔️ Fix: We implemented CSS media queries and a flexible grid layout, ensuring a fully responsive design on different screen sizes. 6️⃣ API Endpoint Not Responding ❌ Bug: At one point, our FastAPI backend stopped responding when making requests for mammogram analysis. ✔️ Fix: This was caused by CORS (Cross-Origin Resource Sharing) restrictions. We fixed it by enabling CORS policies in FastAPI, allowing requests from our frontend to be processed smoothly. 7️⃣ Browser Cache Issues ❌ Bug: Some users reported that they kept seeing old results even after uploading a new image. ✔️ Fix: We added cache-busting techniques in JavaScript and backend headers to ensure that every new request fetches fresh data instead of using the browser’s stored cache. 8️⃣ Inconsistent Accuracy Between Training & Live Testing ❌ Bug: During training, our model achieved high accuracy (75-80%), but in real-world testing, accuracy dropped to ~60%. ✔️ Fix: The problem was due to data imbalance and overfitting. We improved the dataset by: Augmenting real-world mammogram images to improve diversity. Balancing positive and negative cases to avoid bias. Regularizing the model with dropout layers to generalize better on unseen data. After these improvements, we stabilized accuracy at 73% on real-world test cases. Conclusion These challenges helped us improve our system’s stability, speed, and accuracy. By overcoming these bugs, we ensured that our AI-powered breast cancer detection tool provides reliable, fast, and user-friendly results for early detection.