Precise anatomical localization and classification of rib fractures on CT using a convolutional neural network

Area of Science:

• Artificial Intelligence in Medical Imaging
• Radiology and Diagnostic Imaging
• Computational Pathology

Background:

• Rib fractures are common injuries requiring accurate diagnosis and classification.
• Manual detection and classification of rib fractures can be time-consuming and prone to inter-observer variability.
• Automated analysis using deep learning offers potential for improved efficiency and accuracy.

Purpose of the Study:

• To develop a convolutional neural network (CNN) for automated detection, precise anatomical localization, and classification of rib fractures.
• To compare the performance of the developed CNN model against experienced radiologists.
• To evaluate the model's robustness on an external dataset with varying CT scanner parameters.

Main Methods:

• Retrospective collection of 640 rib fracture patient cases with 340,501 annotations from three hospitals.
• Development of a CNN model using RetinaNet architecture with postprocessing for rib localization and result merging.
• Comparative analysis of two model configurations (Model I and Model II) using ROC curves, AUC, precision, recall, and F1-score.

Main Results:

• Model II, incorporating an additional classification model, outperformed Model I in detection and classification.
• High sensitivity for localization: 97.11% for right ribs and 94.87% for left ribs.
• Model II demonstrated superior diagnostic performance on an external dataset and outperformed 5 radiologists in diagnosing fresh and healing fractures, with reduced diagnosis time.

Conclusions:

• The developed CNN model effectively performs automated detection, precise anatomical localization, and classification of rib fractures.
• The AI model shows significant potential to assist radiologists in diagnosing rib fractures, particularly for fresh and healing types.
• The CNN model offers a promising tool for enhancing the efficiency and accuracy of rib fracture assessment in clinical practice.