BioVision: Universal, Cross-Modal Deep Learning for Biomedical Image Analysis

BioVision is a cutting-edge, standalone deep learning software platform that leverages YOLOv11 object detection models and fine-tuned SAM 2.1 for high-performance instance segmentation across diverse biological applications and imaging modalities.

🔬 Key Features

• Multi-Modal Support: Works across ultrasound, X-ray, MRI, electron microscopy, and brightfield microscopy
• Cancer Diagnostics: Accurate classification and segmentation of breast cancer (X-ray mammography) and brain tumors (MRI)
• Comprehensive Organ Segmentation: Supports 34 human organs and 8 mouse tissues from histology datasets
• Fetal Biometry: Precise fetal structure segmentation in ultrasound images for developmental assessment
• Cellular Analysis: White blood cell classification and subcellular structure segmentation (mitochondria)
• User-Friendly Interface: Intuitive GUI with both segmentation and classification workflows

🎯 Applications

Medical Diagnostics

• Breast Cancer Detection: 88.5% classification accuracy, 100% segmentation accuracy
• Brain Tumor Analysis: 99.5% classification accuracy, 97.5% segmentation accuracy
• Fetal Development: 99% accuracy in fetal structure segmentation

Research Applications

• Histopathology: Automated tissue identification across diverse organ systems
• Hematology: 99.4% accuracy in white blood cell classification
• Cellular Biology: Precise mitochondria segmentation in EM images
• Comparative Biology: Cross-species anatomical mapping

🏗️ Architecture

BioVision employs a sophisticated two-stage architecture:

1. Stage 1: Custom-trained YOLOv11 models for robust object detection and region proposal
2. Stage 2: Fine-tuned SAM 2.1 models for precise boundary delineation and instance segmentation

📊 Performance Highlights

• High Accuracy: >88.5% across all applications
• Fast Processing: 4.8-11.9ms per image (depending on modality and resolution)
• Cross-Modal Robustness: Consistent performance across different imaging techniques
• Clinical Precision: Minimal false negatives in cancer detection (0% for breast cancer)

🚀 Getting Started

Requirements

• Operating System: Windows 10 or 11 (64-bit) • Processor: Intel Core i5 or equivalent, or higher • Memory: At least 8 GB RAM (16 GB recommended) • Disk Space: Minimum 500 MB available (Only for the software without models)

💾 Downloads

Complete Software Package

• Google Drive: BioVision Complete Package
  • Pre-trained models for all applications
  • Complete source code
  • Documentation

Pre-trained Models

• YOLOv11 Trained Models: Download All YOLOv11 Models (.pt files)

  • The complete collection of YOLOv11 trained models (.pt files) for all biological applications described in this study
  • Models for cancer detection (breast and brain)
  • Organ segmentation models (34 human + 8 mouse organs)
  • Fetal biometry models
  • White blood cell classification models
  • Cellular structure segmentation models

• SAM2.1 Models: Download All SAM2.1 Models (.pt files)

Individual Components

• Source Code: Available in Google Drive package
• Pre-trained Models: Included in Google Drive package
• Example Datasets: Sample data for testing and validation

🔬 Scientific Impact

BioVision addresses critical challenges in biomedical image analysis by providing:

• Standardized Workflows: Consistent analysis across different imaging modalities
• Clinical Translation: Tools that bridge the gap between research and clinical practice
• Open Science: Fully open-source platform promoting reproducible research
• Multi-Scale Analysis: From subcellular structures to whole organisms

📈 Validation

Our comprehensive evaluation includes:

• Cancer Detection: Outperforming existing methods in brain tumor classification
• Cross-Modal Testing: Validated across 5+ imaging modalities
• Large-Scale Datasets: Tested on thousands of images across diverse biological contexts
• Clinical Relevance: Metrics optimized for real-world medical applications

🤝 Contributing

We welcome contributions from the scientific community!

• Adding new applications
• Improving existing models
• Dataset contributions
• Bug reports and feature requests

📞 Contact

• Corresponding Author: [email protected]
• Department: Genetics and Developmental Biology, Technion Israel Institute of Technology

📄 License

This project is licensed under the CC BY-NC-SA 2.0 License - see the LICENSE file for details.

🙏 Acknowledgments

• Technion Israel Institute of Technology
• The Rappaport Faculty of Medicine and Research Institute
• Google Colab for computational resources
• The open-source community for foundational tools and datasets
• YOLOV11, SAM2.1 and Roboflow for their great efforts in developing the codes and sharing datasets.

---

BioVision: Advancing biomedical research through intelligent image analysis 🔬✨