You will get Breast Cancer AI Diagnosis System with Explainability and Grad-CAM
Project details
Built a full-stack medical AI application combining two independent ML systems into one unified diagnostic decision support tool.
System 1 — Tabular ML: Trained XGBoost, SVM, and Logistic Regression on the Wisconsin Breast Cancer Dataset (569 patients, 30 features). XGBoost achieved 97.4% accuracy and AUC-ROC of 0.994. Integrated SHAP explainability to show which clinical measurements influenced each prediction.
System 2 — Deep Learning: Fine-tuned EfficientNet-B0 via transfer learning on 1,578 breast ultrasound images (benign/malignant/normal), achieving 95.3% validation accuracy. Implemented Grad-CAM from scratch to visualize which regions of the ultrasound the model focused on.
Combined Report: Both models cross-validate each other. If they disagree, the system flags the case as inconclusive and recommends specialist consultation — a critical safety feature for any medical AI tool.
Built with Python, PyTorch, XGBoost, SHAP, Streamlit, OpenCV, scikit-learn.
Delivered as a fully deployed, interactive web application — not just a notebook. Ready to demo live.
System 1 — Tabular ML: Trained XGBoost, SVM, and Logistic Regression on the Wisconsin Breast Cancer Dataset (569 patients, 30 features). XGBoost achieved 97.4% accuracy and AUC-ROC of 0.994. Integrated SHAP explainability to show which clinical measurements influenced each prediction.
System 2 — Deep Learning: Fine-tuned EfficientNet-B0 via transfer learning on 1,578 breast ultrasound images (benign/malignant/normal), achieving 95.3% validation accuracy. Implemented Grad-CAM from scratch to visualize which regions of the ultrasound the model focused on.
Combined Report: Both models cross-validate each other. If they disagree, the system flags the case as inconclusive and recommends specialist consultation — a critical safety feature for any medical AI tool.
Built with Python, PyTorch, XGBoost, SHAP, Streamlit, OpenCV, scikit-learn.
Delivered as a fully deployed, interactive web application — not just a notebook. Ready to demo live.
Programming Languages
PythonCoding Expertise
Performance Optimization, SecurityWhat's included
| Service Tiers |
Starter
$80
|
Standard
$150
|
Advanced
$220
|
|---|---|---|---|
| Delivery Time | 3 days | 5 days | 15 days |
Number of Revisions | 2 | 3 | 5 |
Design Customization | - | - | - |
Content Upload | - | - | - |
Responsive Design | - | - | - |
Source Code |
Frequently asked questions
About Marwa
Machine Learning & Deep Learning Engineer | Computer Vision & NLP
Cairo, Egypt - 8:02 am local time
My work includes:
• A multi-modal breast cancer diagnosis system combining clinical data (XGBoost, 97.4% accuracy) with ultrasound image analysis (EfficientNet-B0, 95.3% accuracy)
• A plant disease detection system with Grad-CAM explainability
• A real-time face recognition attendance system
• Arabic NLP tools for emotion detection and fake news detection
• Sales forecasting dashboards comparing Prophet, LSTM, and ARIMA models
I deploy every project using Streamlit, FastAPI, Vercel, and AWS, so you receive a real working product, not just code.
Certified by Harvard University (Machine Learning & AI), AWS, and IBM (Deep Learning).
Let's turn your data into a real, working AI product. Message me to discuss your project — I'm happy to answer questions before you hire.
Full portfolio: ml-portfolio-frontend-neon.vercel.app
Steps for completing your project
After purchasing the project, send requirements so Marwa can start the project.
Delivery time starts when Marwa receives requirements from you.
Marwa works on your project following the steps below.
Revisions may occur after the delivery date.
Data collection & preprocessing
I collect your clinical data (CSV/Excel) and/or ultrasound images and prepare them for analysis.
Model inference
I run both AI systems — XGBoost tabular model and EfficientNet-B0 image model — on your data to generate predictions.