Advancing Cardio-Oncology
AI for the Early Prediction and Assessment of
Therapy-Induced Cardiotoxicity from Echocardiography
An official MICCAI 2026 challenge dedicated to advancing cardio-oncology. Develop and benchmark your models on a unique, multicentre dataset featuring over 1,900 real-world echocardiograms across 5 longitudinal clinical timepoints.
// Overview
Breast cancer therapy saves lives — but anthracyclines and trastuzumab carry a cardiac cost. Up to 20–30% of patients on anthracyclines and 7–10% on trastuzumab develop measurable cardiac dysfunction, yet the subclinical decline that precedes it is routinely missed until irreversible damage has occurred.
Echocardiography is the monitoring standard across cardio-oncology guidelines, but manual analysis is operator-dependent, time-intensive, and subject to inter-observer variability that erodes the sensitivity needed for early detection in longitudinal surveillance programmes.
The dataset was collected prospectively under the EU Horizon 2020 CARDIOCARE project (Grant No. 945175), capturing real-world clinical heterogeneity across sites — reflecting the population variability that deployed systems will face.
Manual echocardiographic assessment is operator-dependent and highly variable, limiting reliable early detection across longitudinal cancer monitoring programmes.
Existing echocardiography challenges target generic cardiac function. No public dataset pairs therapy-induced cardiotoxicity labels with longitudinal imaging.
From LVEF regression to LV dysfunction classification and early risk prediction — tasks designed around real clinical decision points in cardio-oncology workflows.
// Team
// Get Started
Questions? echorisk.miccai@gmail.com