Sunday, November 30 | 1:20 p.m.-1:30 p.m. | S4-SSMK02-4 | Room E451A
In this session, a deep-learning model developed by a group in China will be presented that could help clinicians diagnose sacroiliitis from standard x-ray images.
While further studies with larger sample sizes will be required, the model demonstrates high diagnostic accuracy, according to presenter Xinyi Meng, PhD, of Tianjin Medical University in Tianjin.
Sacroiliitis is inflammation within the sacroiliac joint, where the lower spine and pelvis meet. Its diagnosis via x-ray typically requires assessments by seasoned radiologists, and unless caught early, spinal rigidity and joint deformities can manifest and the condition becomes irreversible, the researchers noted.
The model was developed using pelvic x-rays from a training set of 465 individuals (930 single sacroiliac joints) and a validation set of 195 individuals (390 single sacroiliac joints). The algorithm was tested using an external test set of 223 individuals (446 single sacroiliac joints).
In the external test set, the model achieved a grading accuracy rate (grades 0 to 4) of 63.9% for sacroiliitis, and its diagnostic accuracy for determining the presence of sacroiliitis reached 90.1%. In addition, with the assistance of the model, the diagnostic accuracy of two junior imaging physicians improved significantly, increasing from 92.5% and 91.1% to 97.2% and 95.3%. Furthermore, the accuracy of the physicians' grading also showed notable improvement, rising from 75% and 74.1% to 88.9% and 80.9%.
With further validation, the model could serve as a valuable computer-aided diagnostic tool for radiologists and rheumatologists, the researchers noted.
