The excellence of Deepnoid solutions, South Korea’s first-generation medical AI (artificial intelligence) specialized company, will be showcased at the world’s largest radiological society.
Deepnoid announced on the 3rd that they will present research results on “AI-based lung nodule diagnostic technology” at the “2024 RSNA (Radiological Society of North America)” held in Chicago from December 1st to 5th, based on U.S. local time. RSNA is the world’s largest radiology and medical association, now in its 109th year since its founding in 1915.
During this event, Deepnoid introduced research results evaluating the diagnostic performance of Deepnoid’s DEEP:LUNG using 455 low-dose chest computed tomography (LDCT) scans from patients who visited outpatient and emergency departments at Busan National University Hospital, Yangsan Busan National University Hospital, and Hwasun Chonnam National University Hospital between January 2019 and July of last year. The research covers tissue, size, malignancy classification, categorization, and localization of nodule positions of lung nodules.
According to Deepnoid, when using DEEP:LUNG, key evaluation metrics recorded a sensitivity of 91.38%, specificity of 93.08%, and an area under the receiver operating characteristic (AUROC) for malignancy classification of 89.62%. Deepnoid explained that an AUROC of over 85% is considered to show significantly good performance.
In particular, the reliability of performance was demonstrated in sensitivity and specificity in the LUNG-RADS category evaluation. For the size measurement of solid nodules and ground-glass opacity nodules, a high precision was maintained within an error range of 2mm and 3mm, respectively.
Choi Woo-sik, the CEO of Deepnoid, stated, “This research has demonstrated that AI can significantly assist medical professionals in the diagnosis and malignancy classification of lung nodules,” and added, “Next year, we plan to expand the application scope of AI solutions to the chest area along with brain disease diagnostic solutions, aiming to provide more comprehensive AI diagnostic support tools in the medical field.”