Full Title: Revolutionizing Liver Tumor Detection: The Novel Unified Inception-U-Net Hybrid Gravitational Optimization Model (UIGO) for Advanced Automated Medical Image Segmentation and Feature Selection

UIGO is a hybrid deep learning framework that integrates:

• Inception modules for multi-scale feature extraction,
• A U-Net-inspired encoder–decoder structure for semantic segmentation,
• Gravitational Optimization for intelligent feature selection.

This model is specifically designed to enhance segmentation performance in liver tumor detection from medical imaging (e.g., CT scans), enabling more accurate and efficient identification of affected regions.

• Encoder: Inception Blocks, Residual Connections, Channel Attention.
• Bottleneck: Multi-path dilated convolutions for contextual learning.
• Decoder: Symmetrical skip connections with upsampling and convolution.
• Optimizer: Gravitational Optimization Algorithm (GOA) to select and refine salient feature representations.

UIGO/ ├── README.md # Project overview and instructions

├── LICENSE # License details

├── requirements.txt # Python dependencies

├── .gitignore # Files and directories to ignore in git

├── data/

│ └── utils.py # Data preprocessing, augmentation utilities

├── models/

│ ├── Segmentation.py # Main hybrid UIGO model

│ ├── Loss.py # Inception, attention, convolutional modules

│ └── GraviationalOptimization.py # Gravitational Optimization Algorithm

│ ├── train.py # Training loop, model fitting

│ ├── loss_functions.py # Loss functions (e.g., Dice loss, focal loss)

│ ├── metrics.py # Evaluation metrics (e.g., IOU, Dice, AUC, precision, recall)

│ └── scheduler.py # Learning rate schedulers and callbacks

├── evaluation/

│ ├── evaluate.py # Testing and final evaluation pipeline

│ └── visualize.py # Prediction visualization and mask overlay

├── utils/

│ └── helpers.py # Logger, seed setting, config reader

├── configs/

│ ├── config.yaml # Hyperparameters, paths, batch size

│ └── dataset.yaml # Dataset splits and data sources

# Clone the repository
git clone https://github.com/yourusername/UIGO.git
cd UIGO

python -m venv venv source venv/bin/activate # or venv\Scripts\activate (Windows)

pip install -r requirements.txt 📊 Evaluation Metrics UIGO is benchmarked using the following metrics:

Dice Coefficient

Intersection over Union (IoU)

Area Under Curve (AUC)

Sensitivity / Specificity

Hausdorff Distance (optional for clinical robustness)

🧪 Datasets The model is evaluated on publicly available liver segmentation datasets:

LiTS (Liver Tumor Segmentation Challenge)

3DIRCADB (3D Image Reconstruction for Comparison of Algorithm Database)

Preprocessing includes resizing, histogram equalization, and data augmentation (flip, rotate, scale).

🚀 Training bash Copy Edit python training/train.py --config configs/config.yaml Parameters like epochs, learning rate, and input size can be modified in the config.yaml file.

🔍 Results UIGO consistently outperforms traditional U-Net variants in terms of segmentation accuracy, robustness, and sensitivity to tumor boundaries. Visualization tools enable side-by-side comparison of predictions with ground truths.

📄 License This project is licensed under the MIT License. See LICENSE for more details.

👨‍🔬 Authors and Contributions This work is developed as part of an advanced medical imaging research initiative. Contributions include model architecture design, gravitational optimization formulation, dataset preparation, and benchmarking.