"Early Detection Saves Lives"
OncoScan leverages the power of deep learning to detect skin cancer using dermatoscopic images and patient metadata for accurate, real-time diagnosis.
OncoScan is a hybrid skin cancer detection system that combines Convolutional Neural Networks (CNNs) for dermatoscopic image analysis with tabular metadata (age, gender, and anatomical site) to enhance classification accuracy.
Current Accuracy - 73-74%
Video Link:- https://drive.google.com/file/d/1zsE8II9oWUNXoy6Bw5OccfVCHBoPtUpL/view (there was error in video we submitted)
made some commit after 4:30 just because of server deploy due to large memory (but now everything is done everything is been fixed)
🎯 The goal is to support early and reliable detection of malignant skin lesions, potentially saving lives through timely diagnosis.
Full Name: Binary Neoplastic Classification with 20,000 dermatoscopic images Images: 20,000 high-resolution dermatoscopic images Metadata:
Patient Age Image Patient Sex Binary Labels: 🟥 Malignant (1) → Includes types like melanoma, basal cell carcinoma, and actinic keratosis 🟩 Benign (0) → Includes all other non-cancerous skin lesion types
🔹 Image Pathway (CNN)
Input: 128×128×3 dermatoscopic image
Conv2D Layers: 32 → 64 → 128 → 256 filters
Each layer: ReLU Activation, Batch Normalization, MaxPooling
Flatten Layer at the end
🔹 Metadata Pathway (Tabular)
Input: 3 features (Age, Sex, Diagnosis_1)
Dense Layer → 32 units with ReLU
Dropout (0.3) for regularization
🔹 Fusion & Output
Concatenation of Image & Metadata branches
Dense → 512 → Dropout (0.5)
Dense → 256 → Dropout (0.3)
Output Layer → 1 unit with Sigmoid activation
Optimizer: Adam
Loss: Binary Crossentropy
ONCOSCAN---SKIN-CANCER-DETECTION/
│
├── app.py # Flask app (main entry)
├── main.py # Additional routing or logic
├── model.py # Model definition/loading
├── data_preprocessing.py # Data preprocessing utilities
├── output.py # Output handling logic (if needed)
│
├── templates/
│ └── index.html # HTML UI template
│
├── static/
│ ├── styles.css # CSS file
│ ├── script.js # JavaScript file
│ ├── logo.png # Web interface logo
│ └── 06-skin-cancer-moles.png # Sample UI image
│
├── README.md # Project documentation
Installation Requirements: Before running the project, install the necessary Python packages: • TensorFlow
• Pandas
• NumPy
• scikit-learn
• matplotlib (optional for visualization)
pip install tensorflow pandas numpy scikit-learn matplotlib
