Feasibility of automatic differential diagnosis of endodontic origin periapical lesions - a pilot study

摘要

Periapical dental diseases (PDD) are one of the most prevalent dental diseases leading to tooth loss and poor quality of life. Differential diagnosis of periapical diseases is critical because the treatment plan depends on the diagnosis. The current diagnosis approach of PDD uses an eyeballing method mainly depending on clinicians' expertise, any lack of which may lead to an inaccurate diagnosis. We developed an advanced image processing tool that can help clinicians for more accurately differential PDD diagnoses that leads to the correct treatment approach. Sixty periapical radiographs were processed by the tool and the differential diagnostic output was compared with manually annotated gold standard dataset. Our tool performed well with a 95% sensitivity, 89% specificity, and 93% accuracy in providing a differential diagnosis. Demonstrating such promising results, further studies should test the accuracy of this tool on a larger dataset get more definitive results for widespread deployment and use.