In this study, researchers focused on leveraging deep learning technology to enhance the diagnosis of axial spondyloarthritis (SpA), specifically by detecting active spinal inflammation through MRI scans. A total of 330 patients diagnosed with axial SpA were involved in the research, from whom STIR sequence MRI scans of the entire spine were collected. The researchers defined spinal inflammation based on the presence of active inflammatory lesions visible in the STIR sequence images.

To facilitate the development of the deep learning model, regions of interest (ROIs) were marked around these inflammatory lesions. The team also created ‘fake-color’ images to simulate adjacent MRI slices for a more comprehensive dataset. This data was then divided, with the majority (images from 270 patients) used to train and validate the deep learning model, while the remaining images (from 60 patients) served as a testing set.

The deep learning model was built using an architecture known as attention-Unet, which is designed to focus on relevant features within the MRI scans for accurate identification of inflammation. After the model was developed, its performance was rigorously tested and compared against the assessments of a general radiologist who was unaware of the algorithm’s findings.

The findings were promising: active inflammatory lesions were identified in 2,891 of the MR images, with the model achieving a sensitivity of 0.80 and specificity of 0.88. The Dice coefficient, a measure of the accuracy of the model’s lesion detection, was 0.55. Additionally, the model achieved an area under the receiver operating characteristic (AUC-ROC) curve of 0.87, indicating a high level of diagnostic accuracy.

In conclusion, the study successfully developed a deep learning algorithm capable of detecting spinal inflammation in axial SpA with accuracy comparable to that of experienced radiologists. This marks a significant step forward in the application of artificial intelligence in medical imaging and diagnosis, particularly in the field of rheumatology.