DeepInfer: Open-Source Deep Learning Deployment Toolkit for Image-Guided Therapy.

Area of Science:

• Artificial Intelligence in Medicine
• Medical Image Analysis
• Computational Biology

Background:

• Deep learning models offer superior performance in medical image analysis compared to traditional methods.
• These models automatically extract hierarchical features, providing flexibility and robustness.
• A significant barrier to clinical adoption is the steep learning curve and technical expertise required for deployment.

Purpose of the Study:

• To introduce DeepInfer, an open-source toolkit designed to simplify the development and deployment of deep learning models.
• To bridge the gap between advanced machine learning in medicine and clinical research evaluation.
• To enable clinical researchers and biomedical engineers to utilize deep learning models without extensive software development.

Main Methods:

• DeepInfer integrates deep learning model development and deployment within the 3D Slicer medical image analysis platform.
• It utilizes a repository of task-specific, pre-trained models accessible via a public registry.
• Users can select and apply trained models to new data without requiring coding or complex configuration.

Main Results:

• DeepInfer facilitates the seamless integration of deep learning into clinical research workflows.
• Demonstrated successful application in prostate segmentation for MRI-guided biopsy.
• Showcased utility in identifying target planes in 3D ultrasound for spinal injections.

Conclusions:

• DeepInfer democratizes the use of advanced deep learning techniques in medical image analysis.
• The toolkit empowers clinical researchers to leverage state-of-the-art AI for improved diagnostic and interventional procedures.
• DeepInfer addresses the practical challenges of deploying deep learning models in clinical research settings.