Using artificial intelligence for distinguishing benign and malignant vertebral compression fractures by computed...

Area of Science:

• Medical Imaging and Artificial Intelligence
• Radiology and Diagnostic Imaging
• Oncology and Bone Diseases

Background:

• Distinguishing benign from malignant vertebral compression fractures (VCFs) is crucial for appropriate patient management.
• Computed tomography (CT) is a primary imaging modality, but differentiating VCFs can be challenging.
• Artificial intelligence (AI) offers potential for automated analysis of medical images.

Purpose of the Study:

• To conduct a scoping review of existing evidence on AI applications for differentiating benign versus malignant VCFs using CT imaging.
• To assess the diagnostic performance and generalizability of AI algorithms in this specific clinical context.

Main Methods:

• Systematic literature search of PubMed/MEDLINE, DOAJ, and ScienceDirect up to October 2025.
• Inclusion of studies employing AI (machine learning, deep learning, hybrid models) on CT images of adult patients with VCFs.
• Focus on studies reporting diagnostic performance metrics and definitive reference standards.

Main Results:

• Six retrospective studies involving 1,767 participants were analyzed.
• AI algorithms demonstrated high diagnostic performance, with AUC scores ranging from 0.76 to 0.99.
• Deep learning and radiomics-based methods showed comparable or complementary performance to radiologists; hybrid models offered synergistic benefits, with some generalizability confirmed in external validation sets.

Conclusions:

• AI-assisted CT evaluation holds significant potential for accurately differentiating benign and malignant VCFs.
• This could potentially reduce unnecessary MRI referrals and expedite initial patient evaluation.
• Further research with robust, multicenter data is necessary to integrate these AI tools into routine clinical practice due to limitations in current retrospective studies.