Professor Park Sung-bae of the Department of Neurosurgery at SMG-SNU Boramae Medical Center, together with a team of researchers from KAIST, announced Tuesday the study results of an AI model that predicts the progression of osteoporotic vertebral fractures using spinal MRI.

For patients with osteoporotic vertebral compression fracture (OVCF), it has been difficult to predict the progression of further spinal damage early based on existing clinical assessments alone. X-rays or CT scans were also inaccurate, so such predictions have had to rely on the experience of medical professionals.

To overcome these limitations, Professor Park's team developed an AI and MRI-based prediction model.

The team retrospectively analyzed the risk of further vertebral fracture progression (VC) in 245 patients with osteoporotic vertebral compression fractures from January 2020 to December 2023 using MRI images and clinical data. Based on data from 200 patients, they then developed an AI-based deep learning model and validated its predictive performance in a test group of 45 patients.

After evaluating the performance of the AI models, the ViT-PMC-LoRA model had the highest prediction accuracy (AUC: 0.8656) compared to other existing models, and the prediction success rate was further improved when augmented prediction techniques were introduced.

“This study has increased the possibility of screening patients with a high risk of vertebral fracture at an early stage,” Professor Park said. “We hope that using this model will improve the diagnostic accuracy of medical professionals and provide practical help in treatment planning.”