AI Demonstrates Potential to Distinguish

Between Simple Osteoporosis and Dangerous Fractures

It has been confirmed that artificial intelligence (AI) can potentially distinguish, using magnetic resonance imaging (MRI), whether a spinal compression fracture—an increasingly common condition with age—was caused simply by osteoporosis or by cancer metastasized to the spine.


(From left) Jaecheol Lee, Professor of Orthopedic Surgery at Soonchunhyang University Seoul Hospital; Junghyun Ahn, Professor of Orthopedic Surgery at Catholic University Bucheon St. Mary's Hospital; Yujin Shin, Professor of Data Science at Catholic University. Soonchunhyang University Seoul Hospital

(From left) Jaecheol Lee, Professor of Orthopedic Surgery at Soonchunhyang University Seoul Hospital; Junghyun Ahn, Professor of Orthopedic Surgery at Catholic University Bucheon St. Mary's Hospital; Yujin Shin, Professor of Data Science at Catholic University. Soonchunhyang University Seoul Hospital

View original image

Soonchunhyang University Seoul Hospital announced on the 24th that a research team led by Professor Jaecheol Lee (corresponding author) from the Department of Orthopedic Surgery at Soonchunhyang University Seoul Hospital, Professor Junghyun Ahn from the Department of Orthopedic Surgery at Catholic University Bucheon St. Mary's Hospital, and Professor Yujin Shin from the Department of Data Science at Catholic University has applied deep learning technology to spinal MRIs, which are commonly performed at hospitals, in an effort to solve this issue.


As people age, their bones weaken, making them more susceptible to compression fractures of the spine. Most often, this is due to osteoporosis, but in rarer cases, it can be caused by cancer from other organs spreading (metastasizing) to the spine and weakening the bone.


Although the symptoms appear similar on the surface, the treatment approaches for these two causes are completely different. If osteoporosis is the cause, treatment involves osteoporosis medications and procedures to stabilize the spine. However, if cancer metastasis is the cause, it is required to identify the original cancer and provide entirely different treatments such as chemotherapy or radiotherapy.


The issue arises when a fracture caused by cancer is incorrectly regarded as a simple osteoporosis fracture—delaying the crucial detection and treatment of cancer. For this reason, it is extremely important to distinguish between these two types of fractures accurately.


The research team first trained AI to independently detect the location of abnormal vertebrae on MRI images and then to distinguish whether those abnormalities were fractures due to osteoporosis or cancer metastasis. To achieve this, they analyzed MRIs from 2,165 patients with either spinal compression fractures or spinal metastases and assessed the accuracy of the AI across 27,543 vertebral segments. Additionally, multiple AI models were trained on the same set of images to compare their accuracy and determine which model performed the best.


As a result, the AI model that produced the best performance showed about a 91% probability of being correct when diagnosing cancer metastasis as the cause, and about a 93% detection rate among actual cases of metastatic spinal fractures. This model demonstrated the most balanced performance overall, including accurate identification of the affected location. Notably, some models exhibited particularly excellent capability in not missing metastatic spinal fractures, suggesting future potential for use in screening examinations.


Professor Jaecheol Lee stated, "This study is significant in that it confirms the possibility of using deep learning technology on routine spinal MRIs to help medical professionals identify lesions that require special attention and to differentiate between cancer and non-cancer cases. If applied as an auxiliary tool—not replacing doctors, but acting as an additional check where something might otherwise be missed—it could assist in delivering faster and more accurate diagnoses and treatments."



This research was published in the international journal for spinal medicine, The Spine Journal.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing