This work shows how Imagen’s FDA-cleared software for detecting fractures (FractureDetect) in musculoskeletal radiographs improves clinicians’ accuracy at diagnosing fractures. Both clinicians with extensive training in musculoskeletal imaging and clinicians with limited training were significantly more accurate when assisted by our deep-learning system.
At Imagen, we are working to change how diagnostics are delivered. The following are peer-reviewed publications written by us about the current and future state of imaging diagnostics.
In this work, Imagen shows the effectiveness of Chest-CAD, our FDA-cleared software that analyzes chest radiograph studies using artificial intelligence. Our software highlighted a suspicious part of an X-ray missed through unaided interpretation, thus showing the promise of the additional information provided by our artificial intelligence software.
Imagen developed a deep-learning system for detecting fractures across the musculoskeletal system. In a multi-site study, the deep-learning system accurately identified fractures throughout the adult musculoskeletal system. The results were used, in part, to support FDA clearance of Imagen’s software (FractureDetect) to assist clinicians in detecting fractures for a selected subset of the anatomic regions reported in this paper.
Imagen developed a deep neural network to detect and localize fractures in wrist radiographs. A controlled experiment showed that the average clinician had a 47% relative reduction in their misinterpretation rate using the deep-learning model. This significant improvement in diagnostic accuracy shows that deep-learning methods can substantially enhance patient care.