Professor Lee Seok-hyun of Hallim University's Gangnam Sacred Heart Hospital won the Korean Society of Nuclear Medicine 'Young Researcher Award'
Nov 14, 2024
Lee Seok-hyun, a professor of radiology at Hallim University's Gangnam Sacred Heart Hospital, won the Young Researcher Award at the 63rd Korean Nuclear Medical Association Fall Conference held at the SC Convention Center (Korea Science and Technology Center) in Gangnam-gu, Seoul from the 1st to the 2nd.
Professor Lee Seok-hyun was honored with the presentation of 'Comparison of Diagnostic Performance of CNN and Transformer Models Using Grad-CAM for Bone Transition Diagnosis in Bone Scans'.
Bone scanning is a test that shows the bone activation area as an image, and is performed to check for inflammation, damage, or cancer metastasis of the bone. In particular, bone scans are mainly performed for prostate cancer and breast cancer patients with relatively common bone metastases. This is because the patient's burden is low because the bones of the whole body can be checked at once at a lower cost than CT or MRI.
Recently, research on artificial intelligence (AI) models has been actively conducted in various medical images such as X-rays and endoscopes, but there are still few artificial intelligence studies on bone scans. In particular, not many studies have been conducted on whether state-of-the-art artificial intelligence models with significantly improved performance, such as Transformer models and ConvNeXt, are useful in actual medical imaging diagnosis.
In response, Professor Lee Seok-hyun's research team (Professor Kim Dong-woo of the Department of Nuclear Medicine at Hallym University Sacred Heart Hospital, Professor Son Hye-joo of the Department of Nuclear Medicine at Dankook University Hospital, and Park Se-hyun of Hallym University Medical School) compared the results of applying several latest artificial intelligence models to bone scanning and suggested the possibility that artificial intelligence models could be useful for diagnosis. In particular, this study drew attention in that it cross-verified the performance of artificial intelligence models in bone scanning with external data as well as internal hospital data.
The research team compared the diagnostic performance of several artificial intelligence models with a total of 6,175 patients, including 4,694 patients at Hallim University Gangnam Sacred Heart Hospital and 1,481 patients at Hallim University Sacred Heart Hospital. Among AI models, ResNet, the most widely applied in the medical imaging field, Transformer models used in ChatGPT, and ConvNeXt, which improved ResNet, were used in the study. As a result, ResNet was 63% sensitive and 90% specificity when diagnosing bone metastasis in bone scan, while ConvNeXt was able to diagnose bone metastasis more accurately with 79% sensitivity and 100% specificity.
Professor Lee Seok-Hyun believes that the latest AI models, such as ConvNeXt, can be widely used in several medical images, including bone scanning, based on research results"We will continue to conduct various studies on the application of the latest technology so that patients can be diagnosed with medical imaging more quickly and accurately in the future."."
Professor Lee Seok-hyun was honored with the presentation of 'Comparison of Diagnostic Performance of CNN and Transformer Models Using Grad-CAM for Bone Transition Diagnosis in Bone Scans'.
Bone scanning is a test that shows the bone activation area as an image, and is performed to check for inflammation, damage, or cancer metastasis of the bone. In particular, bone scans are mainly performed for prostate cancer and breast cancer patients with relatively common bone metastases. This is because the patient's burden is low because the bones of the whole body can be checked at once at a lower cost than CT or MRI.
Recently, research on artificial intelligence (AI) models has been actively conducted in various medical images such as X-rays and endoscopes, but there are still few artificial intelligence studies on bone scans. In particular, not many studies have been conducted on whether state-of-the-art artificial intelligence models with significantly improved performance, such as Transformer models and ConvNeXt, are useful in actual medical imaging diagnosis.
In response, Professor Lee Seok-hyun's research team (Professor Kim Dong-woo of the Department of Nuclear Medicine at Hallym University Sacred Heart Hospital, Professor Son Hye-joo of the Department of Nuclear Medicine at Dankook University Hospital, and Park Se-hyun of Hallym University Medical School) compared the results of applying several latest artificial intelligence models to bone scanning and suggested the possibility that artificial intelligence models could be useful for diagnosis. In particular, this study drew attention in that it cross-verified the performance of artificial intelligence models in bone scanning with external data as well as internal hospital data.
The research team compared the diagnostic performance of several artificial intelligence models with a total of 6,175 patients, including 4,694 patients at Hallim University Gangnam Sacred Heart Hospital and 1,481 patients at Hallim University Sacred Heart Hospital. Among AI models, ResNet, the most widely applied in the medical imaging field, Transformer models used in ChatGPT, and ConvNeXt, which improved ResNet, were used in the study. As a result, ResNet was 63% sensitive and 90% specificity when diagnosing bone metastasis in bone scan, while ConvNeXt was able to diagnose bone metastasis more accurately with 79% sensitivity and 100% specificity.
Professor Lee Seok-Hyun believes that the latest AI models, such as ConvNeXt, can be widely used in several medical images, including bone scanning, based on research results"We will continue to conduct various studies on the application of the latest technology so that patients can be diagnosed with medical imaging more quickly and accurately in the future."."
|
bellho@sportschosun.com