Trustworthy medical AI starts with better data

This session is part of a series of webinars organized by national member societies of the International Society for Telemedicine & eHealth (ISfTeH). In this edition, the Czech Society for Biomedical Engineering and Medical Informatics discusses insights into the trustworthy use of AI in healthcare. The growing use of data‑driven methods in healthcare makes data quality a critical concern. Even the most advanced analytical or AI techniques cannot overcome poor‑quality inputs. The principle "garbage in, garbage out" remains highly relevant in modern medicine, where inaccurate or biased data may lead to flawed models, misleading conclusions, and potentially harmful clinical decisions. Medical data quality depends on completeness, consistency, accuracy, timeliness, and validity, yet healthcare data are often collected from heterogeneous sources under varying clinical conditions. Missing values, inconsistent definitions, and measurement differences can undermine reliability, while human error, device limitations, or technical failures may directly affect patient care and research outcomes. Two major challenges are noise and bias, which reflects systematic distortion caused by unrepresentative populations, unequal data collection, or hidden confounders. Ensuring high‑quality data therefore requires careful design of the data collection process, including standardized protocols, clear variable definitions, calibrated equipment, quality‑control procedures, and balanced datasets. Ultimately, trustworthy, representative, and responsibly collected data are not only a technical foundation for healthcare AI but also an ethical necessity. Speaker: Lenka Lhotska, PhD Associate Professor, Department of Natural Sciences, Faculty of Biomedical Engineering & Department of Cognitive Systems and Neurosciences, Czech Institute of Informatics, Robotics and Cybernetics Czech Technical University in Prague Scientific Secretary, Czech Society for Biomedical Engineering and Medical Informatics