Compression fractures of the spine, common in the elderly and osteoporosis patients, require early detection for effective treatment and complication prevention. Although CT is widely used for diagnosis, thick-slice CT images often obscure fine structural details, making accurate diagnosis difficult. Recent advances in AI enable generating high-resolution images from low-resolution inputs. Diffusion models have recently emerged as an alternative to address the limitations of generative adversarial networks (GAN)-based models, which often suffer from training instability and inconsistent results. In this paper [1], the authors propose a 3D conditional diffusion model-based slice interpolation method to improve vertebral height estimation and compression fracture detection from thick-slice CT images (Figure 1). The Quantitative Morphometry (QM) method automatically detects and classifies fracture types based on height ratios (Figure 2). Compared to GAN-based methods, the diffusion model successfully captured diverse fracture morphologies (Figure 3), with a sensitivity of 93.0% and a specificity of 97.3%. In the future, it is expected that this method will enhance diagnostic accuracy for spinal compression fractures and have potential for early detection of other abnormalities.
[1] Kudo, A., Kitamura, Y., Suzuki, Y., Tomiyama, N., Hori, M. (2026). 3D Super-Resolution for Enhancing Compression Fracture Detection in Thick-Slice CT: Diffusion Models vs GANs. In: Fernandez, V., Wiesner, D., Zuo, L., Casamitjana, A., Remedios, S.W. (eds) Simulation and Synthesis in Medical Imaging. SASHIMI 2025. Lecture Notes in Computer Science, vol 16085. Springer, Cham.
DOI:https://doi.org/10.1007/978-3-032-05573-6_4
CAUTION：This is Fujifilm Global Website. Fujifilm makes no representation that products on this website are commercially available in all countries. Approved uses of products vary by country and region. Specifications and appearance of products are subject to change without notice.