CINA-VCF

CINA-VCF is an artificial intelligence-driven software that analyzes CT scans of the chest and abdomen to detect potential Vertebral Compression Fractures (VCFs) in patients aged 50 and over, operating by processing DICOM images using deep learning models trained to identify VCFs between T1-L5. Its purpose is to assist medical specialists by flagging suspected cases for prioritized review, thereby enhancing workflow triage in bone health settings without serving as a diagnostic tool itself.

CINA-VCF is a radiological computer-aided triage and notification software designed to identify suspected Vertebral Compression Fractures (VCFs) in adult patients aged 50 and over undergoing non-enhanced or contrast-enhanced CT scans of the chest and/or abdomen. The system employs deep learning algorithms, trained on a dataset of 886 CT series from multiple centers, to analyze images and detect VCFs at the T1-L5 vertebral levels. Its methodology involves receiving DICOM images, processing them to identify potential fractures while also locating the spine, identifying vertebral bodies, and excluding vertebrae with prior surgical interventions like vertebroplasty or orthopedic material. The software integrates with existing medical image communication systems, such as the CINA Platform, to flag suspected positive cases on a worklist application, providing a compressed, unannotated preview image (marked "not for diagnostic use") to facilitate earlier triage and prioritization by medical specialists without altering the original images or replacing standard diagnostic procedures. The performance of CINA-VCF was validated through a retrospective, multicenter, multinational, blinded study on 474 clinical cases, comparing its findings against a ground truth established by three board-certified radiologists. The primary endpoint, the Area Under the Receiver Operating Characteristic curve (ROC AUC), was 0.974, exceeding the predefined performance goal of 0.95. The software demonstrated a sensitivity of 95.2% and a specificity of 92.9%, with an overall accuracy of 93.7%. A key secondary outcome was the time-to-notification; for true positive cases, CINA-VCF achieved a mean notification time of 21.7 seconds, significantly faster than the predicate device. These results indicate the software's capability to effectively identify suspected VCFs and expedite their review, aiding in workflow efficiency for bone health assessments.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the CINA-VCF algorithm differentiate between varying grades of vertebral compression fractures (e.g., mild, moderate, severe), and does this grading influence the triage notification or prioritization?; What is the performance impact on CINA-VCF when encountering CT scans with significant artifacts or in patients with atypical spinal anatomy not extensively represented in the training dataset of 886 series?; How does the system account for confounding factors such as degenerative changes, scoliosis, or other spinal pathologies that might mimic or obscure vertebral compression fractures?; What mechanisms are in place to ensure the algorithm's performance remains robust over time, considering potential shifts in CT imaging technology or patient populations?; Beyond worklist prioritization, what studies have been conducted or are planned to assess the real-world clinical impact of CINA-VCF on time-to-diagnosis, intervention rates, or patient outcomes for VCFs?