LiverAI Pipeline — v1.2

This project implements a unified liver analysis system combining CT-based tumor detection and ultrasound-based fatty liver classification.

Features

• CT scan tumor detection (slice-based)
• Fatty liver classification (ultrasound-based)
• Combined prediction pipeline
• Morphology analysis (tumor size, liver area, ratios)
• Saved prediction outputs (PNG + JSON)

Project Structure

src/ct_pipeline/ → CT pipeline scripts
src/fatty_pipeline/ → fatty liver training notebook
src/combined/ → unified prediction system
data/ct_liver/processed/ → generated slices and masks
data/ct_liver/outputs/ → predictions and metrics
models/ → trained models

Models

• CT Model → Normal vs Tumor classification
• Fatty Liver Model → NAFLD vs Non-NAFLD

Dataset

The CT dataset used is Task03_Liver from the Medical Segmentation Decathlon.

Download from: https://medicaldecathlon.com/

After downloading, place it in:

data/ct_liver/raw/Task03_Liver/

Setup

1. Create environment: py -3.11 -m venv venv
   venv\Scripts\activate

2. Install dependencies: python -m pip install -r requirements.txt

Combined Prediction

python src/combined/combined_predict.py <image_path>

Outputs:

• Combined prediction image
• JSON report including:
  • Fatty liver result
  • CT tumor result
  • Morphology metrics (if available)
  • Overall health status

Outputs

• Prediction visualizations (PNG)
• Structured JSON results
• Morphology measurements

Notes

• Both models use standardized input size (224x224)
• CT and ultrasound modalities are processed independently and combined at output level
• Morphology analysis is available only when corresponding mask files exist
• Raw dataset (~26GB) is not included in the repository