Summary: A Hong Kong-based team has developed the Co-Plane Attention across MRI Sequences deep learning model to classify 12 knee abnormalities from multi-sequence MRI, improving diagnostic accuracy, efficiency, and aligning with clinical standards while bridging expertise gaps in radiology.

Key Takeaways

  • The Co-Plane Attention across MRI Sequences (CoPAS) model enhances knee MRI diagnostic accuracy by effectively classifying 12 common knee abnormalities using multi-sequence imaging data.
  • The AI model demonstrated performance comparable to radiologists, outperforming junior radiologists in diagnostic accuracy while improving the efficiency and accuracy of senior radiologists.
  • The study highlights the potential of artificial intelligence to bridge expertise gaps in healthcare, aligning the model’s decision-making process with clinical standards and improving patient care.

———————————————————————————————————————————————————————————————————

Multi-sequence knee MRI is an essential diagnostic method for identifying knee pathologies, but its interpretation is time-consuming and demands high expertise. A research team from the School of Engineering at the Hong Kong University of Science and Technology, in collaboration with the Third Affiliated Hospital of Southern Medical University in China, has developed a novel deep learning model to classify 12 common knee abnormalities, enhancing diagnostic accuracy and efficiency.

CoPAS Enhances Knee MRI

The study, published in Nature Communications under the title “Learning Co-Plane Attention Across MRI Sequences for Diagnosing Twelve Types of Knee Abnormalities”, describes the use of the Co-Plane Attention across MRI Sequences (CoPAS) model. By analyzing T1-weighted, T2-weighted, and proton density-weighted MRI sequences from sagittal, coronal, and axial planes, and integrating data from 1,748 patients, the model effectively captured intensity variations and decoupled spatial features to identify abnormalities with high accuracy. Arthroscopy, the gold standard for diagnosing knee abnormalities, was used to validate the findings.

“Accurately diagnosing knee abnormalities is crucial for customizing treatment plans and improving patients’ quality of life,” says assistant professor Chen Hao, PhD, who led the research. “This innovative CoPAS model demonstrates diagnostic performance comparable to that of radiologists. It is particularly beneficial in bridging the gap between less experienced and senior doctors.”

Improving Radiologist Diagnoses

In simulated clinical testing, radiologists first diagnosed MRI scans independently. After a washout period, they repeated the process using the model’s output for reference. Results showed the model outperformed junior radiologists in accuracy while assisting senior radiologists in improving their diagnoses. Furthermore, a comparative analysis found the model’s decision-making process aligned with clinical standards, reflecting its ability to emulate radiologists’ reasoning.

“Our findings underscore the promise of artificial intelligence in healthcare, highlighting its potential to identify and validate new clinical insights,” Chen adds.