Soombit AI Secures MFDS Approval for AI Radiology Software

Subscribe to Our Newsletter

Success! Now Check Your Email

To complete Subscribe, click the confirmation link in your inbox. If it doesn’t arrive within 3 minutes, check your spam folder.

Ok, Thanks
Soombit AI Secures MFDS Approval for AI Radiology Software
Source: Soombit AI

Soombit AI Secures MFDS Approval for AI Radiology Software

South Korean firm Soombit AI receives Class 3 MFDS approval for AIRead-CXR, a generative AI tool that drafts preliminary chest X-ray reports.

Philip Lee profile image
by Philip Lee

Seoul, South Korea - Medical artificial intelligence company Soombit AI has received Class 3 approval from the Ministry of Food and Drug Safety (MFDS) for AIRead-CXR, a generative AI software that drafts preliminary radiology reports from chest X-rays, the company announced.

According to Soombit AI, this is the first global regulatory approval for a generative AI-based Software as a Medical Device (SaMD) that operates as a standalone product for medical image analysis and draft report generation.

The approval follows clinical trials conducted at Seoul National University Hospital and Seoul National University Bundang Hospital, which used over 1,000 chest X-ray datasets to assess the software's validity and safety.

The AIRead-CXR model was trained on approximately 14 million domestic and international data points. It processes both posteroanterior (PA) and anteroposterior (AP) chest X-rays.

The software identifies 57 findings and diseases, including cavities, nodules, and infiltrations, and quantifies severity using standardized terminology.

The system synthesizes these imaging findings to provide differential diagnosis information in the impression section of the report.

Published clinical data indicate that thoracic radiologists accepted AIRead-CXR's draft reports at an average rate of 85%, compared to 66.9% for the MedGemma model.

The software recorded a hallucination rate of 0.3%, versus MedGemma's 9.7%.

The clinical efficacy of the software has been documented in three papers published in the journal Radiology and one in the American Journal of Roentgenology (AJR).

Before receiving approval, Soombit AI conducted a global web demonstration of AIRead-CXR for 2 years to incorporate feedback from medical specialists into the development process.

Philip Lee profile image
by Philip Lee

Subscribe to The Pickool

Success! Now Check Your Email

To complete Subscribe, click the confirmation link in your inbox. If it doesn’t arrive within 3 minutes, check your spam folder.

Ok, Thanks

Read More