Introduction to the course objectives, structure and expected learning outcomes. Short discussion on how AI is already influencing healthcare and medical practice.

Introduction to Artificial Intelligence, Machine Learning, Deep Learning and Generative AI. Rule-based systems versus data-driven learning, supervised and unsupervised learning, classification and regression, illustrated through healthcare examples.

Overview of healthcare data types, including clinical data, medical images, biomedical signals and EHR information. Introduction to features, labels, ground truth and the AI lifecycle: training, validation, testing and deployment.

Introduction to confusion matrix, accuracy, sensitivity, specificity, precision, F1-score, ROC-AUC and calibration. Discussion of false positives, false negatives, class imbalance and why technical performance does not automatically mean clinical usefulness.

Participants analyse simplified healthcare AI scenarios, interpret performance metrics and discuss whether a model would be appropriate for use in a specific clinical context.

Review of the main concepts and preparation for Day 2.

Short recap of Day 1 and introduction to the main areas in which AI is currently applied in healthcare.

Overview of AI applications for image classification, detection, segmentation and enhancement. Introduction to AI analysis of ECG, EEG and other biomedical signals, including event detection, monitoring and early warning.

How AI can support continuous monitoring outside the hospital through wearable devices and sensors. Discussion of data collection, remote monitoring, alert generation and integration into clinical workflows.

Introduction to Clinical Decision Support Systems (CDSS), including alerts, risk scores and recommendations. Rule-based versus AI-based decision support and the importance of providing the right information to the right person at the right moment.

Use of Generative AI, NLP and speech recognition for clinical notes, discharge summaries, referrals, information extraction and medical record summarization. Discussion of hallucinations, omissions and the need for clinician verification.

Examples of AI for patient flow, discharge planning, appointment management and patient communication. Discussion of chatbots, patient navigation and when AI should escalate to a healthcare professional.

Human-in-the-loop, automation bias, fairness, privacy, accountability and traceability. Participants analyse an AI-generated clinical document or patient-facing interaction and identify risks and required safeguards.

Presentation of the practical challenge, formation of teams and explanation of the final deliverable. Participants may work on a healthcare problem they already know or select one from a predefined set of challenges.

Each team defines the problem it wants to solve, the target users or patients, the current workflow, why the problem is relevant and where AI could potentially add value.

Teams define the core elements of their proposed solution: required data, type of AI functionality, expected output, intended user and position of the AI system within the healthcare workflow.

Teams define how the solution would be evaluated, relevant performance indicators, possible risks and limitations, privacy and data considerations, and where human oversight is required.

Trainers work with each team to challenge assumptions, clarify the AI use case and improve the proposed solution.

Each team presents its healthcare AI concept in a short 5–7 minute pitch, covering the problem, proposed AI solution, required data, workflow integration, expected benefits, evaluation approach, risks, limitations and human oversight.

Final feedback, discussion of the proposed solutions and key takeaways from the three-day course.