Comprehensive Summary
This study aimed to evaluate the ability of a deep learning hip MRI protocol called Compressed Sense Artificial Intelligence (CSAI) to detect labral abnormalities in patients with femoroacetabular impingement syndrome (FAIS). Early detection is critical to reduce the risk of hip osteoarthritis, but the current common imaging, magnetic resonance (MR), cannot detect more subtle labral abnormalities. CSAI and Compressed Sense (CS) systems enhance image quality and reduce imaging times, allowing for more rapid and accurate diagnoses. 32 patients with symptomatic FAIS and who had 3 prior MRI were selected and imaged with both CS and CSAI tools. Four qualified independent evaluators assessed the produced images for presence or absence of labral abnormalities in specific zones. Hip arthroscopy findings were used as a reference to determine absence or presence of such defects. For labral abnormalities, both CS and CSAI achieved highly accurate results, with high sensitivity (97-100%) and perfect specificity, at 100%. There was found to be no significant difference between CS and CSAI performance. As for cartilage lesions, both performed poorly, however CSAI performed marginally better than CS in sensitivity (42% vs 37%) and specificity (44% vs 39%). Lowest performance for both was found in the anteroinferior and posterior zones of the acetabulum and the inferior and posterior zones of the femur, where both CS and CSAI scored less than 6% sensitivity. Overall, these tools are strong when detecting labral abnormalities, however still require refinement in more anatomically challenging areas.
Outcomes and Implications
While deep learning imaging methods are still imperfect, this article poises CS and CSAI as imaging methods of the future in labral pathology. These tools can speed up imaging times and boost accuracy over traditional MRI methods, allowing for a much more streamlined and efficient imaging process. With continued improvement, these tools could allow surgeons to produce a more complete plan of operation before entering the operating room, allowing surgery to become more efficient as well. However, some time and refinement are needed for tools such as CS and CSAI before they can be fully adopted into clinical decision making.