Update README with comprehensive documentation for all new features

README.md

@@ -12,20 +12,32 @@ license: apache-2.0
 
 # NeuroSAM 3: Medical Image Segmentation
 
-A medical image segmentation application using SAM 3 (Segment Anything Model 3) for DICOM file analysis.
+A comprehensive medical image segmentation application using SAM 3 (Segment Anything Model 3) for DICOM file analysis with advanced features.
 
-## Features
+## ✨ Features
 
+### Core Features
 - 🧠 **SAM 3 Integration**: Uses the latest Segment Anything Model 3 for medical image segmentation
 - 📁 **DICOM Support**: Process CT and MRI DICOM files
 - 🖼️ **Image Formats**: Supports DICOM (.dcm), PNG, and JPG files
 - 🎯 **Text Prompts**: Describe what you want to segment (e.g., "brain", "tumor", "skull")
 - ⚙️ **Windowing Strategies**: Optimized windowing presets for CT images
-- 🎨 **Visualization**: Overlay segmentation masks on medical images
-- 📊 **Interactive Slice Viewer**: Scroll through multiple slices from the same subject
+
+### Interactive Features
+- 📊 **Interactive Slice Viewer**: Scroll through multiple slices with auto-play
 - 🔍 **Subject Detection**: Automatically groups images by patient/subject ID
+- ▶️ **Auto-play**: Automatically animate through slices
 - 📈 **Ground Truth Comparison**: Compare SAM 3 results with expert annotations (Dice score, IoU)
 
+### Enhanced Features (NEW!)
+- ✨ **Enhanced Single Image**: Brightness/contrast controls, colormap options, transparency adjustment
+- ✨ **Enhanced Batch Processing**: Process multiple images with ZIP download
+- 🎯 **Point/Box Prompts**: Interactive point-click and bounding box-based segmentation
+- 📊 **ROI Statistics**: Detailed statistics (area, intensity, centroid, bounding box)
+- 📥 **NIFTI Export**: Export segmentation masks to NIFTI format for medical imaging software
+- 💾 **Annotation Save/Load**: Save and reload annotations for later use
+- 🎭 **Multi-Mask Output**: Generate multiple mask candidates with confidence scores
+
 ## Usage
 
 ### Single Image Processing
@@ -40,7 +52,32 @@ A medical image segmentation application using SAM 3 (Segment Anything Model 3)
 2. Click "🔍 Detect Subjects" to auto-group by patient ID
 3. Select a subject from the dropdown
 4. Click "Process All Slices"
-5. Use the slider or navigation buttons to scroll through slices
+5. Use the slider, navigation buttons, or ▶️ Auto-play to scroll through slices
+
+### Enhanced Single Image (with Image Adjustments)
+1. Upload an image
+2. Adjust brightness and contrast sliders
+3. Choose colormap and transparency for mask overlay
+4. Click "Segment with Progress"
+5. Download result using the download button
+
+### Point/Box Prompts
+1. Upload an image
+2. **Point Prompt**: Enter X, Y coordinates and click "Segment at Point"
+3. **Box Prompt**: Enter bounding box coordinates (X1, Y1, X2, Y2) and click "Segment in Box"
+
+### ROI Statistics & Export
+1. Process an image using "Process & Calculate Statistics"
+2. View detailed ROI statistics (area, intensity, centroid)
+3. Export to NIFTI format for use in medical imaging software
+4. Save annotation for later use
+5. Load previously saved annotations
+
+### Multi-Mask Output
+1. Upload an image and enter prompt
+2. Choose number of masks to generate (1-5)
+3. Click "Generate Multiple Masks"
+4. Compare mask candidates with confidence scores
 
 ### Compare with Ground Truth
 1. Upload a medical image and its ground truth mask
@@ -54,6 +91,8 @@ A medical image segmentation application using SAM 3 (Segment Anything Model 3)
 - Gradio
 - PyDICOM
 - Transformers (with SAM 3 support)
+- nibabel (for NIFTI export)
+- scipy (for ROI statistics)
 
 ## Model
 
@@ -67,6 +106,21 @@ This app uses the SAM 3 model from Facebook/Meta. You need:
 - UniqueData/dicom-brain-dataset (Hugging Face) - MRI Brain scans
 - The Cancer Imaging Archive (TCIA) - Various medical imaging
 - Imaging Data Commons (IDC) - Large collection of DICOM files
+- Kaggle Brain MRI datasets - PNG/JPG format images
+
+## Tabs Overview
+
+| Tab | Description |
+|-----|-------------|
+| Single Image | Basic single image segmentation |
+| Interactive Slice Viewer | Browse through multiple slices with auto-play |
+| Gallery View | View all segmentations in a grid |
+| Compare with Ground Truth | Evaluate with Dice/IoU metrics |
+| ✨ Enhanced Single Image | Advanced controls with download |
+| ✨ Enhanced Batch Processing | Batch processing with ZIP export |
+| 🎯 Point/Box Prompts | Interactive region selection |
+| 📊 ROI Statistics & Export | Statistics + NIFTI/annotation export |
+| 🎭 Multi-Mask Output | Multiple mask candidates |
 
 ## License