🧠 Real-Time Brain Tumor Detection in Intraoperative Ultrasound

This repository contains the official implementation of the method described in:

Cepeda S, Esteban-Sinovas O, Romero R, et al. “Real-time brain tumor detection in intraoperative ultrasound: From model training to deployment in the operating room.” Computers in Biology and Medicine. 2025;193:110481. doi:10.1016/j.compbiomed.2025.110481

https://www.sciencedirect.com/science/article/pii/S0010482525008327

---

🚀 Overview

This repository contains the code and model checkpoint for real-time detection of brain tumors using intraoperative ultrasound (ioUS) and the YOLO11 architecture.
The system was prospectively tested in the operating room to evaluate feasibility and integration into the surgical workflow.

⚠️ Disclaimer: This software is intended for research purposes only and is not approved for clinical use.

---

🔍 Highlights

• Real-time brain tumor detection with 2D ioUS images
• Trained on a multicenter dataset of 1,732 annotated images
• Prospective validation in 20 brain tumor surgeries
• Achieves mAP@50 = 0.95 and FPS = 34.16 with YOLO11s
• Designed for efficient deployment on low-cost consumer-grade hardware

---

📁 Repository Structure

📦 RealTime-BrainTumor-Detection-YOLO11
├── inference_real_time.py     # Real-time inference script (optimized for OR)
├── inference_off_line_save.py # Offline inference from a recorded ultrasound video, saving the results with overlaid predictions
├── checkpoints(detect_s.pt)    # Trained model weights
├── LICENSE
└── README.md

---

💡 Getting Started

1. Clone the repo

git clone https://github.com/your-username/RealTime-BrainTumor-Detection-YOLO11.git
cd RealTime-BrainTumor-Detection-YOLO11

2. Install dependencies

conda create -n yolobrain python=3.10
conda activate yolobrain
pip install -r requirements.txt

3. Run real-time inference

🖥️ Setup Example

• Connect your ultrasound machine to your computer using a USB video capture card.
• Identify the correct camera index (e.g., 2) for the virtual camera.
• Place the trained YOLO11 model in the same directory (e.g., detect-s.pt).

Run the script:

python inference_real_time.py

---

🎞️ inference_off_line_save.py

Performs offline detection on a recorded ultrasound video. It processes each frame, applies our trained model, and saves a new video with overlaid predictions.

✅ Features

• Frame-by-frame processing with prediction overlay
• Saves annotated video in .mp4 format
• GPU acceleration supported
• Optional live display during processing

🛠️ Setup

Edit the script to define:

video_path = 'path/to/your_ultrasound_video.mp4'
output_video_path = 'path/to/save_annotated_video.mp4'

Then run:

python inference_off_line_save.py

🧪 Model Performance Summary (YOLO11s)

Metric	Value
mAP@50	0.95
mAP@50-95	0.65
Latency	24.9 ms
FPS	34.16
Model size	18.3 MB
Parameters	9.43 M

---

🧠 About the Project

This work was developed by the GEIBAC Group,
Specialized Group in Biomedical Imaging and Computational Analysis at IBioVALL – University of Valladolid.

---

🔒 License

This repository is licensed under a non-commercial research-only license.
See the LICENCE file for terms.

Contact us for licensing or collaboration requests:
📧 [email protected]

---

📝 Citation

If you use this code or model in your research, please cite:

@article{CEPEDA2025110481,
title = {Real-time brain tumor detection in intraoperative ultrasound: From model training to deployment in the operating room},
journal = {Computers in Biology and Medicine},
volume = {193},
pages = {110481},
year = {2025},
issn = {0010-4825},
doi = {https://doi.org/10.1016/j.compbiomed.2025.110481},
url = {https://www.sciencedirect.com/science/article/pii/S0010482525008327},
author = {Santiago Cepeda and Olga Esteban-Sinovas and Roberto Romero and Vikas Singh and Prakash Shett and Aliasgar Moiyadi and Ilyess Zemmoura and Giuseppe Roberto Giammalva and Massimiliano {Del Bene} and Arianna Barbotti and Francesco DiMeco and Timothy R. West and Brian V. Nahed and Ignacio Arrese and Roberto Hornero and Rosario Sarabia},
}

@software{yolo11_ultralytics,
  author = {Glenn Jocher and Jing Qiu},
  title = {Ultralytics YOLO11},
  version = {11.0.0},
  year = {2024},
  url = {https://github.com/ultralytics/ultralytics},
  orcid = {0000-0001-5950-6979, 0000-0002-7603-6750, 0000-0003-3783-7069},
  license = {AGPL-3.0}
}

---

📽️ Demo

▶️ Demo video from OR deployment