Automatic Detection of Spinal Pathologies Based on MRI Scans

This study introduces an automated system for detecting spinal pathologies in MRI scans, aiming to enhance efficiency and accuracy in medical diagnostics. By leveraging deep learning, the tool can assist radiologists in identifying and classifying pathological vertebrae, potentially improving diagnostic workflows.

MRI scans are essential for diagnosing various lumbar spine conditions, including infections, inflammations, and carcinomas. Currently, radiologists interpret these scans, which is both time-consuming and subject to variability. With the increasing volume of MRI scans, there is a growing need for efficient solutions.

The primary goal is to develop an automated system to identify pathological vertebrae in lumbar spine MRI scans, streamlining diagnostic workflows and assisting radiologists by pinpointing areas of concern.

The method employs a RetinaNet algorithm for detection and a ConvNeXt model for classification with three different approaches (2D with majority vote, 2D with LightGBM, and 3D images). It is unique for its combination of advanced detection and classification models and its thorough exploration of multiple classification methods.

• Detection: The object detection model achieved an Intersection-Over-Union (IOU) score of 0.67 for pathological vertebrae.
• Classification: The 2D-based model with LightGBM post-processing demonstrated effective performance, achieving 95.38% accuracy, 86.67% precision, 92.86% recall, and an F1-score of 89.66%
• Efficiency: Automating the detection and classification process can reduce the time radiologists spend on each scan, allowing them to focus on critical cases and improve overall workflow efficiency.
• Accuracy: Improved diagnostic accuracy can reduce the risk of misdiagnosis and ensure better patient outcomes.
• Scalability: The system’s integration into existing diagnostic workflows is seamless, making it scalable across various medical facilities without extensive retraining of staff.

This automated detection tool represents the integration of AI in medical diagnostics, aiming to enhance efficiency and accuracy in the interpretation of lumbar spine MRI scans. As the healthcare industry continues to evolve, such advancements can help meet the increasing demand for quality medical care.

Future Directions
Future research will focus on expanding the model’s capabilities to include a broader range of spinal pathologies and integrating additional MRI sequences to further enhance diagnostic accuracy.

Sign up for
our events

    Life Science