Skip to content

More than a decade of peer-reviewed evidence.

A continuous clinical research program validating Cardio Explorer®.

Validated across the cardiovascular risk spectrum

Cardio Explorer® was developed, refined, and validated through retrospective diagnostic accuracy studies spanning algorithm development, model lock, and intended-use validation. A prospective implementation study is now evaluating clinical adoption, usability, and impact on patient management in routine care.

Obstructive CAD

Anatomical endpoint
University Hospital Basel

Basel I

Initial algorithm development and internal validation

First feasibility cohort: machine-learned pattern recognition trained on patients referred for invasive coronary angiography.

LURIC Study

LURIC

Algorithm refinement and lock in a high-risk population

Locked model evaluated on independent validation sets within a high cardiovascular-risk cohort.

Maastricht University

Maastricht

Pivotal evaluation in the intended-use population

Validated in the intended-use population: low-to-intermediate-risk outpatients referred for CAD workup.

iGES

AOK Niedersachsen

Prospective implementation in primary care

Multicentre controlled study evaluating Cardio Explorer® acceptance, usability, and impact on diagnostic pathways in general practice.

Myocardial ischemia

Functional endpoint
University Hospital Basel

Basel II

Validation of the algorithm for ischemia prediction

Prospective single-centre cohort validating the memetic pattern-based algorithm (MPA) against Rubidium-82 PET myocardial perfusion imaging.

Publication

Original research · Open access

Artificial intelligence to improve ischemia prediction in Rubidium Positron Emission Tomography — a validation study

Frey SM, Bakula A, Tsirkin A, et al. EPMA J. 2023;14:631–643.

doi:10.1007/s13167-023-00341-5

Download PDF(opens in a new tab)

Study design

Design
Prospective single-centre cohort; validation of the memetic pattern-based algorithm (MPA) for ischemia prediction
Population
Consecutive patients referred for Rubidium-82 PET myocardial perfusion imaging (ischemia work-up for suspected/known CAD)
Sample size
2,417 validation
Setting
Tertiary centre, University Hospital Basel, Switzerland (Jul 2018 – Feb 2022)
Reference standard
Rubidium-82 PET myocardial perfusion imaging (rest / pharmacologic stress with adenosine or regadenoson)
Ischemia definition
Summed difference score (SDS) ≥ 2

Consistent diagnostic performance across populations

Performance of the current CE-marked algorithm (Cardio Explorer® v1.6.0) in three validation cohorts: how well it discriminates, how safely it rules out obstructive CAD, and how reliably it rules it in at the thresholds used in the clinical pathway.

Area under the ROC curve

Bar chart. AUC by cohort: Basel I 0.835, LURIC 0.875, Maastricht 0.878.

Source: Exploris Health clinical performance data for Cardio Explorer® v1.6.0. AUC and the 15% threshold values are documented in the Instructions for Use v1.6.0, Table 1.

Scientific evidence behind Cardio Explorer®

Journal articles, congress abstracts and registered trials on the Memetic Pattern Algorithm and Cardio Explorer®, newest first: from the 2014 letter that introduced the algorithm to the 2026 comparison with the ESC 2024 model.

  1. 2026
    Congress abstract

    Comparison of a novel scientific model of the memetic pattern based (MPA v3) and the ESC 2024 RF-CL algorithm for the detection of coronary artery disease: a retrospective multicentre study

    Buesser AJ, Frey S, Kugler V, et al. European Heart Journal – Digital Health 7(Suppl 1):ztaf143.046.

    Retrospective multicentre comparison of MPA v3 with the ESC 2024 risk-factor-weighted clinical likelihood (RF-CL) model for detecting coronary artery disease. Validation cohort n = 413.

    Authors
    Buesser AJ, Frey S, Kugler V, Haaf P, Zellweger MJ
    DOI
    10.1093/ehjdh/ztaf143.046
    Summary

    Retrospective multicentre study comparing the third-generation memetic pattern based algorithm (MPA v3) with the ESC 2024 risk-factor-weighted clinical likelihood (RF-CL) model for the detection of coronary artery disease. In the validation cohort (n = 413), mean CAD prevalence was 50%. The authors conclude that MPA v3 outperformed the ESC 2024 RF-CL algorithm across key diagnostic metrics.

  2. 2025
    Review, in German

    Coronary heart disease — a German paradox. Using AI-based technology to solve a relevant healthcare problem

    Remppis BA, Zellweger MJ, Kelle S, et al. Herzschrittmachertherapie + Elektrophysiologie 37(1):57–68.

    Review of coronary artery disease diagnostics in Germany and of AI-based pre-test probability, drawing on the Cardio Explorer® validation studies and a 2024 budget impact analysis.

    Authors
    Remppis BA, Zellweger MJ, Kelle S, Leber A, Weißer M, Heinl PV, Freudenberg LS, Rischpler C, Emrich T, Mildenberger P, Helms T
    DOI
    10.1007/s00399-025-01094-6
    Abstract

    The German healthcare system is facing challenges in diagnosing coronary artery disease (CAD). These include high mortality rates, even when advanced medical technology is used, and an excessive number of coronary angiographies. A key issue is that current guideline models inaccurately estimate pretest probability (PTP). However, accurately determining PTP is crucial for efficiently guiding the diagnostic pathway and ensuring cost-effectiveness. Artificial intelligence (AI)-based systems offer a solution for precisely determining a personalized PTP. Approved medical AI devices calculate a personalized PTP using laboratory values and medical history. These systems achieve high accuracy (area under the curve 0.87), improve patient selection, avoid unnecessary diagnostics for low-risk patients, and demonstrate significant cost savings. AI-based PTP is essential for improving guideline-based CAD diagnostics and utilizing resources more efficiently.

  3. 2024
    Registered trial

    Risk assessment for chronic coronary heart disease by general practitioners: Using the Cardio Explorer in primary care

    ISRCTN Registry, ISRCTN67912485.

    Prospectively registered, controlled, non-randomised study of Cardio Explorer® use by general practitioners in Lower Saxony, sponsored by AOK. Completed March 2026 with 484 participants; results not yet published.

    DOI
    10.1186/ISRCTN67912485
    Registry summary

    Background and study aims. Coronary heart disease is still considered a widespread disease in Germany despite slightly declining mortality rates. The challenges in the early detection of patients with coronary heart disease in combination with developments in healthcare policy (e.g. limited specialist resources combined with long waiting times) show the considerable need to improve risk assessment in primary care. This study investigates the implementation of an artificial intelligence-based algorithm, the Cardio Explorer. The main goal is to investigate the level of acceptance, usability and trust in the Cardio Explorer by general practitioners.

  4. 2023
    Journal article

    Artificial intelligence to improve ischemia prediction in Rubidium Positron Emission Tomography — a validation study

    Frey SM, Bakula A, Tsirkin A, et al. EPMA Journal 14:631–643.

    Validation in 2,417 consecutive patients referred for Rubidium-82 PET: the memetic pattern-based algorithm predicted ischemia more accurately than the ESC and ACC pre-test probability models (AUC 0.758 vs 0.585–0.673).

    Authors
    Frey SM, Bakula A, Tsirkin A, Vasilchenko V, Ruff P, Oehri C, Amrein MF, Huré G, Rumora K, Schäfer I, Caobelli F, Haaf P, Mueller CE, Remppis BA, Brunner-La Rocca HP, Zellweger MJ
    DOI
    10.1007/s13167-023-00341-5
    Abstract

    Background. Patients are referred to functional coronary artery disease (CAD) testing based on their pre-test probability (PTP) to search for myocardial ischemia. The recommended prediction tools incorporate three variables (symptoms, age, sex) and are easy to use, but have a limited diagnostic accuracy. Hence, a substantial proportion of non-invasive functional tests reveal no myocardial ischemia, leading to unnecessary radiation exposure and costs. Therefore, preselection of patients before ischemia testing needs to be improved using a more predictive and personalised approach.

    Aims. Using multiple variables (symptoms, vitals, ECG, biomarkers), artificial intelligence–based tools can provide a detailed and individualised profile of each patient. This could improve PTP assessment and provide a more personalised diagnostic approach in the framework of predictive, preventive and personalised medicine (PPPM).

    Methods. Consecutive patients (n = 2417) referred for Rubidium-82 positron emission tomography were evaluated. PTP was calculated using the ESC 2013/2019 and ACC 2012/2021 guidelines, and a memetic pattern–based algorithm (MPA) was applied incorporating symptoms, vitals, ECG and biomarkers. Five PTP categories from very low to very high PTP were defined (i.e., < 5%, 5–15%, 15–50%, 50–85%, > 85%). Ischemia was defined as summed difference score (SDS) ≥ 2.

    Results. Ischemia was present in 37.1%. The MPA model was most accurate to predict ischemia (AUC: 0.758, p < 0.001 compared to ESC 2013, 0.661; ESC 2019, 0.673; ACC 2012, 0.585; ACC 2021, 0.667). Using the < 5% threshold, the MPA’s sensitivity and negative predictive value to rule out ischemia were 99.1% and 96.4%, respectively. The model allocated patients more evenly across PTP categories, reduced the proportion of patients in the intermediate (15–85%) range by 29% (ACC 2012)–51% (ESC 2019), and was the only tool to correctly predict ischemia prevalence in the very low PTP category.

    Conclusion. The MPA model enhanced ischemia testing according to the PPPM framework: 1) The MPA model improved individual prediction of ischemia significantly and could safely exclude ischemia based on readily available variables without advanced testing (“predictive”). 2) It reduced the proportion of patients in the intermediate PTP range. Therefore, it could be used as a gatekeeper to prevent patients from further unnecessary downstream testing, radiation exposure and costs (“preventive”). 3) Consequently, the MPA model could transform ischemia testing towards a more personalised diagnostic algorithm (“personalised”).

  5. 2022
    Journal article

    Use of artificial intelligence to assess the risk of coronary artery disease without additional (non-invasive) testing: validation in a low-risk to intermediate-risk outpatient clinic cohort

    Eurlings CGMJ, Bektas S, Sanders-van Wijk S, et al. BMJ Open 12:e055170.

    Validation in 696 low-to-intermediate-risk outpatients at Maastricht University Medical Centre (16.2% CAD prevalence). AUC 0.87, compared with 0.80 for pre-test probability and 0.82 for the CAD consortium clinical model.

    Authors
    Eurlings CGMJ, Bektas S, Sanders-van Wijk S, Tsirkin A, Vasilchenko V, Meex SJR, Failer M, Oehri C, Ruff P, Zellweger MJ, Brunner-La Rocca HP
    DOI
    10.1136/bmjopen-2021-055170
    Abstract

    Objectives. Predicting the presence or absence of coronary artery disease (CAD) is clinically important. Pretest probability (PTP) and CAD consortium clinical (CAD2) model and risk scores used in the guidelines are not sufficiently accurate as the only guidance for applying invasive testing or discharging a patient. Artificial intelligence without the need of additional non-invasive testing is not yet used in this context, as previous results of the model are promising, but available in high-risk population only. Still, validation in low-risk patients, which is clinically most relevant, is lacking.

    Design. Retrospective cohort study.

    Setting. Secondary outpatient clinic care in one Dutch academic hospital.

    Participants. We included 696 patients referred from primary care for further testing regarding the presence or absence of CAD. The results were compared with PTP and CAD2 using receiver operating characteristic (ROC) curves (area under the curve (AUC)). CAD was defined by a coronary stenosis >50% in at least one coronary vessel in invasive coronary or CT angiography, or having a coronary event within 6 months.

    Outcome measures. The first cohort validating the memetic pattern-based algorithm (MPA) model developed in two high-risk populations in a low-risk to intermediate-risk cohort to improve risk stratification for non-invasive diagnosis of the presence or absence of CAD.

    Results. The population contained 49% male, average age was 65.6±12.6 years. 16.2% had CAD. The AUCs of the MPA model, the PTP and the CAD2 were 0.87, 0.80, and 0.82, respectively. Applying the MPA model resulted in possible discharge of 67.7% of the patients with an acceptable CAD rate of 4.2%.

    Conclusions. In this low-risk to intermediate-risk population, the MPA model provides a good risk stratification of presence or absence of CAD with a better ROC compared with traditional risk scores. The results are promising but need prospective confirmation.

  6. 2018
    Journal article

    A new non-invasive diagnostic tool in coronary artery disease: artificial intelligence as an essential element of predictive, preventive, and personalized medicine

    Zellweger MJ, Tsirkin A, Vasilchenko V, et al. EPMA Journal 9(3):235–247.

    Validation in 987 patients of the Ludwigshafen Risk and Cardiovascular Health Study (LURIC). The optimised MPA reached AUC 0.87 in the test population; the Framingham Risk Score reached 0.69 in the training population.

    Authors
    Zellweger MJ, Tsirkin A, Vasilchenko V, Failer M, Dressel A, Kleber ME, Ruff P, März W
    DOI
    10.1007/s13167-018-0142-x
    Abstract

    Background. Known coronary artery disease (CAD) risk scores (e.g., Framingham) estimate the CAD-related event risk rather than presence/absence of CAD. Artificial intelligence (AI) is rarely used in this context.

    Aims. This study aims to evaluate the diagnostic power of AI (memetic pattern-based algorithm (MPA)) in CAD and to expand its applicability to a broader patient population.

    Methods and results. Nine hundred eighty-seven patients of the Ludwigshafen Risk and Cardiovascular Health Study (LURIC) were divided into a training (n = 493) and a test population (n = 494). They were evaluated by the Basel MPA. The “training population” was further used to expand and optimize the Basel MPA, and after modifications, a final validation was carried out on the “test population.” The results were compared with the Framingham Risk Score (FRS) using receiver operating curves (ROC; area-under-the-curve (AUC)). Of the 987 LURIC patients, 71% were male, age 62 ± 11 years and 68% had documented CAD. AUC of Framingham and BASEL MPA to diagnose CAD in “LURIC training” were 0.69 and 0.80, respectively. AUC of the optimized MPA in the training and test cohort were 0.88 and 0.87, respectively. The positive predictive values (PPV) of the optimized MPA for exclusion of CAD in “training” and “test” were 98 and 95%, respectively. The PPV of MPA for identification of CAD was 93 and 94%, respectively.

    Conclusions. The successful use of the MPA approach has been demonstrated in a broad-risk spectrum of patients undergoing CAD evaluation, as an element of predictive, preventive, personalized medicine, and may be used instead of further non-invasive diagnostic procedures.

  7. 2014
    Letter to the editor

    A new memetic pattern based algorithm to diagnose/exclude coronary artery disease

    Zellweger MJ, Brinkert M, Bucher U, et al. International Journal of Cardiology 174(1):184–186.

    Letter introducing the memetic pattern based algorithm (MPA). In 373 patients, the MPA reached AUC 0.82 in the validation cohort, compared with 0.74 for the Framingham Risk Score.

    Authors
    Zellweger MJ, Brinkert M, Bucher U, Tsirkin A, Ruff P, Pfisterer ME
    DOI
    10.1016/j.ijcard.2014.03.184
    Summary

    The memetic pattern based algorithm (MPA) was built in a training population of 245 patients evaluated for coronary artery disease by routine testing and validated in 128 separate patients, with invasive coronary angiography as the reference. Of the 373 patients, 66% were male, aged 65 ± 10 years, and 66% had CAD. The AUC of the MPA and the Framingham Risk Score was 0.79 and 0.66 in the training population and 0.82 and 0.74 in the validation cohort.