Cleerly ISCHEMIA

Cleerly ISCHEMIA is a machine learning-based software designed as an adjunctive diagnostic tool to non-invasively assess the likelihood of coronary ischemia. It analyzes quantitative CCTA data, including atherosclerosis and stenosis metrics, to provide a binary ischemia index benchmarked against invasive FFR, aiming to support clinical decision-making for patients with suspected coronary artery disease.

Cleerly ISCHEMIA is an automated, machine learning-based decision support software designed to aid in the diagnosis of ischemia associated with coronary artery disease (CAD). Functioning as an add-on module to Cleerly Labs, it analyzes Coronary Computed Tomography Angiography (CCTA) images to quantify atherosclerosis, stenosis, and vascular morphology, thereby determining the likely presence or absence of coronary vessel ischemia. The software employs a "locked" algorithm, meaning it does not continuously learn, and processes eligible vessel data through pre-processing and post-processing modules. Its output, the Cleerly ISCHEMIA Index (CII), is a binary indicator (positive/negative for likely ischemia) benchmarked against an invasive Fractional Flow Reserve (FFR) threshold of ≤0.80, and results are visually displayed, with vessels indicating likely ischemia illuminated in red, to non-invasively support the functional evaluation of stable symptomatic CAD patients. The clinical performance of Cleerly ISCHEMIA was validated primarily through the CREDENCE trial, a prospective, multicenter study involving 612 stable patients with suspected CAD, using invasive FFR as the reference standard. In this trial, the software demonstrated a per-vessel territory sensitivity of 75.9% and a specificity of 83.4% for its primary endpoint. Further performance data, including pooled results from US and international studies, showed combined per-vessel sensitivity and specificity of 76.2% and 85.2% respectively, with a negative predictive value (NPV) of 90.5%. On a per-patient basis, the pooled data indicated a sensitivity of 86.6% and a specificity of 69.8%, with an NPV of 84.6%. Non-clinical testing also verified the software's risk assessment, performance, and cybersecurity.

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 'locked' nature of the Cleerly ISCHEMIA algorithm affect its adaptability to evolving CCTA imaging technologies or new understandings of ischemia predictors over time?; What specific quantitative measures of atherosclerosis, stenosis, and vascular morphology are weighted most heavily by the algorithm in determining the Cleerly ISCHEMIA Index, and are these weightings transparent to the user?; Given the difference in specificity between per-vessel (85.2% pooled) and per-patient (69.8% pooled) analyses, what are the clinical implications for potential over- or under-treatment based on the unit of analysis?; What were the exclusion criteria for vessels or studies in the pre-processing module, and how might these exclusions impact the generalizability of the tool in a diverse patient population with complex coronary anatomy?; How does the performance of Cleerly ISCHEMIA compare to other non-invasive ischemia assessment modalities, such as stress echocardiography or SPECT MPI, particularly in subgroups of patients defined by disease severity or comorbidities?