Anatomy-centred deep learning improves generalisability and progression prediction in radiographic sacroiliitis detection.

PURPOSE: To examine whether incorporating anatomy-centred deep learning can improve generalisability and enable prediction of disease progression.

METHODS

This retrospective multicentre study included conventional pelvic radiographs of four different patient cohorts focusing on axial spondyloarthritis collected at university and community hospitals. The first cohort, which consisted of 1483 radiographs, was split into training (n=1261) and validation (n=222) sets. The other cohorts comprising 436, 340 and 163 patients, respectively, were used as independent test datasets. For the second cohort, follow-up data of 311 patients was used to examine progression prediction capabilities. Two neural networks were trained, one on images cropped to the bounding box of the sacroiliac joints (anatomy-centred) and the other one on full radiographs. The performance of the models was compared using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity and specificity.

RESULTS

On the three test datasets, the standard model achieved AUC scores of 0.853, 0.817, 0.947, with an accuracy of 0.770, 0.724, 0.850. Whereas the anatomy-centred model achieved AUC scores of 0.899, 0.846, 0.957, with an accuracy of 0.821, 0.744, 0.906, respectively. The patients who were identified as high risk by the anatomy-centred model had an OR of 2.16 (95% CI 1.19, 3.86) for having progression of radiographic sacroiliitis within 2 years.

CONCLUSION

Anatomy-centred deep learning can improve the generalisability of models in detecting radiographic sacroiliitis. The model is published as fully open source alongside this study.