This project implements a three-tier web application designed for the classification of lung histology image patches (histopathological images) into Malignant (Cancerous: Adenocarcinoma/Squamous Cell Carcinoma) or Benign (Non-Cancerous).
The application uses a PyTorch deep learning model served by a FastAPI backend and accessed via an Angular frontend.
-
Frontend: Angular (Standalone Components, TypeScript, HttpClient)
-
Backend/API: FastAPI (Python)
-
Machine Learning: PyTorch (Transfer Learning with EfficientNet-B0)
-
Data Handling: Uploads handled via FastAPI, model weights loaded via model.py.
The model was trained using Transfer Learning on the LC25000 Lung Histology Image Dataset, utilizing an EfficientNet-B0 architecture.
The plot below shows the training performance, where the model successfully converged to near-zero loss on the validation set, indicating robust generalization.
To run this project, you need two separate terminals—one for the backend (Python) and one for the frontend (Angular).
NOTE: All Python commands must be executed within the activated venv_311 environment.
- Node.js/NPM: Must be installed in your WSL/Ubuntu environment.
- Python 3.11+: Must be installed (ideally in a virtual environment).
- Data: The histology_classifier_final.pth model weights file must be present in the backend/ directory.
Step 1: Start the Backend (FastAPI)
The backend must be started first to ensure the frontend can connect to the running PyTorch model.
-
Open your WSL/Ubuntu terminal.
-
Navigate to the project root (LUNG-CANCER-DETECTION-CODEJAM/).
-
Activate the Python environment: source venv_311/Scripts/activate
-
Start the FastAPI server: python -m uvicorn backend.main:app --reload
--> Wait for the console to display: INFO: Application startup complete.
Step 2: Start the Frontend (Angular)
-
You will need a second terminal (still running WSL/Ubuntu) for the Angular development server.
-
Open a new WSL/Ubuntu terminal.
-
Navigate to the Angular project directory (e.g., cd frontend/ui).
-
Run the serve command: ng serve
--> Wait for the console to display: Compiled successfully. Angular Live Development Server is listening on localhost:4200
Step 3: Access the Application
Open your web browser and navigate to:
You can now upload a histology image patch (JPEG or PNG) and see the prediction returned by the PyTorch model running in the FastAPI backend!