Lunit INSIGHT CXR

The study developed a deep learning algorithm, DLAD-10, to detect 10 common abnormalities (Nodule, Consolidation, Pneumothorax, Pleural Effusion, Atelectasis, Pneumoperitoneum, Cardiomegaly, Mediastinal Widening, Calcification, Fibrosis)  on chest radiographs and assessed its impact on diagnostic accuracy, reporting timeliness, and workflow efficiency. DLAD-10 demonstrated high performance with area under the receiver operating characteristic curve values ranging from 0.895 to 1.00 and improved the detection of critical and urgent abnormalities compared to radiologists alone. The use of DLAD-10 significantly reduced the time-to-report and interpretation time for critical and urgent cases, enhancing overall diagnostic efficiency.