metadata
license: mit
pipeline_tag: image-segmentation
tags:
- medical
- biology
CT Heart Segmentation Model
Model Information:
- Architecture: U-Net
- Task: Binary segmentation of heart in CT scans
- Dataset: Heart CT Dataset
- Input Size: 224×224 RGB images
Performance Metrics:
- Best Dice Score: 0.9479
- Best IoU Score: 0.9014
Usage:
from shifaa.vision import VisionModelFactory
model = VisionModelFactory.create_model(
model_type="segmentation",
model_name="CT_Heart"
)
results = model.run("heart_ct.png", show_image=True)
image = results["image"]
mask = results["predicted_mask"]
Sample Results:
Architecture Details:
- Encoder: 4 downsampling blocks (Conv → BatchNorm → ReLU → Dropout)
- Bottleneck: Deepest convolutional block
- Decoder: 4 upsampling blocks with skip connections
- Output: 1 channel with sigmoid activation
Preprocessing:
- Random horizontal flip
- Random rotation ±15°
- Random brightness & contrast adjustment
- Normalize and convert to tensor
Training Details:
- Loss Function: Combined Dice Loss + BCE Loss
- Optimizer: Adam (lr=0.001, weight_decay=1e-5)
- Batch Size: 8
- Epochs: 100 (with early stopping)