Deep Learning Body Region Classification of MRI and CT Examinations.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Accurate anatomical labeling of medical images is crucial for diagnosis and treatment planning.
• Current methods for body region identification can be labor-intensive and prone to error.
• Deep learning offers a potential solution for automated and accurate anatomical classification.

Purpose of the Study:

• To develop and evaluate a deep learning model for identifying body regions in computed tomography (CT) and magnetic resonance (MR) axial images.
• To assess the model's performance across diverse acquisition protocols, manufacturers, and patient demographics.
• To achieve high accuracy in classifying seventeen CT and eighteen MRI body regions covering the entire human body.

Main Methods:

• A convolutional neural network (CNN)-based classifier was developed for pixel-based anatomical labeling.
• Three retrospective datasets comprising 2891 CT and 3339 MRI cases from 27 institutions were used for training, validation, and testing.
• The model's sensitivity and specificity were evaluated against various factors including patient age, sex, institution, scanner, contrast, slice thickness, and imaging sequences/kernels.

Main Results:

• The deep learning model achieved high image-level weighted sensitivity of 92.5% for CT and 92.3% for MRI.
• Weighted specificity reached 99.4% for CT and 99.2% for MRI.
• The model demonstrated robust performance across diverse patient populations and imaging parameters, including lower and upper extremities.

Conclusions:

• Deep learning models, specifically CNNs, can accurately classify CT and MR images by body region.
• The developed AI model shows high performance in anatomical labeling, covering the entire human body.
• This technology has the potential to enhance efficiency and accuracy in radiological image analysis.