… is an AI project based on data from the Study of Osteoporotic Fracture (SOF), with hundreds of publications the most widely published study in the field of osteoporosis. In SOFIA we will pursue two aims:
(i) Assessing the predictive power of deep neural network assessment of pelvic and hand radiographs for incidence of hip and other osteoporotic fractures and
(ii) Artificial intelligence based identification of features on pelvic radiographs indicative of enhanced hip fracture risk.
A major concern regarding AI is its nature as a black box: it is not clear how the network arrived at its conclusion. This may hinder acceptance. Therefore, we will also aim to increase the interpretability of AI based fracture risk prediction. We plan to get insight into the AI black box by using visualization methods such as heat maps and by deep generative methods to better understand the model learned by the AI.
SOFIA is a cooperation of the i²LAB with the Institute of Computer Science of the CAU and US researchers at UCSF, Harvard Medical School and other US universities and research centers.