Ultrasound AI in clinical use from June... Tracks signs with deep learning, increases diagnosis rate Indicates malignancy level with colors Imaging AI, primary differentiation of cancer findings... "Helpful for Korean women with dense breast tissue" Forecast by Professor Lee Kwang-hee of Samsung Changwon Hospital
● Real-time Ultrasound Diagnosis AI Developed Domestically
Professor Lee stated that the global incidence of breast cancer is increasing by 10-20% annually. He noted that the incidence rate is rapidly increasing, especially among those under 35. Professor Lee mentioned, "There is even a term called 'Young Breast Cancer' for this case." Breast cancer can be well-treated with surgery and procedures if detected early. The remaining challenge is to enhance the level of diagnosis.
Breast cancer diagnosis heavily relies on imaging examinations. According to Professor Lee, one reason breast cancer diagnosis is challenging for Korean women is the prevalence of dense breasts. This is when the proportion of glandular tissue is higher than fatty tissue in the breast. About 60-70% of Korean women have dense breasts. Professor Lee explained, "In ultrasound, the shadows of glandular tissue and breast cancer often appear similar. While breast cancer detection is easier when there is more fatty tissue, it is difficult to distinguish when there is more glandular tissue, as in dense breasts. Therefore, small findings are sometimes missed in imaging diagnosis."
For these reasons, mammography and breast ultrasound examinations have been conducted concurrently in Korea. Recently, AI that enhances the efficacy of both has been developed and applied clinically.
"Mammography AI was introduced three years ago. However, there were challenges in developing ultrasound examination AI that confirms suspicious areas in real-time due to frequent time lag issues. Last year, a domestic AI capable of following in real-time, whether for 10 minutes or an hour, was developed and released this year. It has a definite advantage in diagnosing small cancer signs that are not detected early."
Samsung Changwon Hospital has been using the real-time breast ultrasound AI solution 'CADAI-B,' developed by Beamworks, a company founded by experts in medical imaging and AI, since June. A test of CADAI-B's breast cancer detection sensitivity on 800 people at Yonsei University Severance Hospital showed a 95% rate.
● Analyzing Cancer Malignancy and Progression Probability to Assist Physicians
Mammography AI was introduced in Korea in 2022 and is being utilized by many medical institutions. AI initially distinguishes breast cancer findings such as calcification, structural distortion, and high-risk masses, and predicts malignancy. Professor Lee noted, "In some Northern European countries, it is legally required for two physicians to interpret and sign off on mammography. However, with the development of AI, these countries now legally recognize the interpretation by one physician and AI. AI can be seen as playing the role of an assistant physician."
At Samsung Changwon Hospital, AI diagnostic costs are not being charged initially. It is still uncertain whether AI diagnostic costs will be covered by health insurance benefits in the future. For now, it is recommended before examinations. The hospital is investing at this stage to expand its base. Professor Lee expressed confidence, "If a system where 2D and 3D X-rays are taken, breast ultrasound is performed, and AI diagnosis is included is well established, there will be no interpretation errors. Gradually, Korea could be recognized as a world-leading diagnostic nation for breast cancer."
Above all, Professor Lee hopes that people with breast cancer risk factors accurately understand AI diagnosis. He stated, "I hope they become subjects of awareness rather than objects who only trust the physician's words." He added, "Physicians may resist AI, believing it is not yet perfect, and less experienced physicians may trust it vaguely. It is time to consider the accurate understanding and appropriate clinical application of evolving AI."
Yu Jae-young
AI-translated with ChatGPT. Provided as is; original Korean text prevails.
ⓒ dongA.com. All rights reserved. Reproduction, redistribution, or use for AI training prohibited.
About Dong-A Ilbo