Automatic Body Region Classification in CT Scans Using Deep Learning.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Accurate anatomical region classification in computed tomography (CT) is crucial for medical imaging analysis.
• Existing methods may face challenges with variations in imaging protocols and patient demographics.

Purpose of the Study:

• To demonstrate the high performance of deep learning (DL) algorithms in classifying whole-body parts from CT scans.
• To evaluate the effectiveness of DL for automated body region classification in diverse CT datasets.

Main Methods:

• A deep learning model was trained on 5485 anonymized CT scans from 45 health centers.
• Scans were classified into six distinct body region classes (chest, abdomen, pelvis, and combinations).
• The dataset was divided into training (3290), validation (1097), and testing (1098) sets.

Main Results:

• The DL model achieved high performance metrics: 97.53% accuracy, 97.56% precision, 97.6% recall, and 97.56% F1-score.
• The model demonstrated robustness across varied acquisition protocols and patient demographics.
• The classification covered six whole-body regions, including chest, abdomen, and pelvis.

Conclusions:

• Deep learning models show significant potential for automating body region classification in CT scans.
• This approach can enhance diagnostic efficiency and consistency in clinical radiology workflows.
• The study highlights the strength of DL in annotating CT images with high accuracy.