MultiMeDIA Lab at CinC 2026

From the 20th to 23rd September the 53rd Computing in Cardiology (CinC) conference took place in Madrid, Spain. Held annually since 1974, CinC is an international scientific conference focused on research applying computing in clinical cardiology and cardiovascular physiology. At this year's conference, the MultiMeDIA Lab delivered 3 poster presentations and 1 oral presentation, including receiving the Best Poster Award. Fourth year undergraduate student Charlotte Richardson was invited for an oral presentation, and MRes student Adrian McIntosh and DPhil students Thalia Seale and Alexander Sharp were invited for poster presentations.

Charlotte's work, titled "Multi-Modal Supervised Contrastive Learning for Myocardial Infarction Classification", presents a new multi-modal deep learning framework for classifying myocardial infarction. Two variational autoencoder-based encoders separately encode ECG and MRI information, and three fusion strategies are explored to combine these latent representations and project them to a shared space. Results across fusion strategies show that multi-modal fusion of ECG and cardiac MRI consistently outperform single-modality methods. Thalia, second author of the paper, delivered this presentation during Session 2: Cardiovascular Models, on 22 September from 11:45-12:00.

Alex's paper, titled “Reconstructing Conduction Velocity Maps from Sparse Measurements in Atrial Fibrillation,” presents a statistical framework for reconstructing dense conduction velocity maps of the left atrium from sparse clinical measurements. Using a statistical appearance model trained on charge density mapping data from 49 patients with persistent atrial fibrillation, the work shows that sampling approximately 20-30% of the atrial surface is sufficient to recover much of the spatial information in the full conduction velocity map. Alex presented this work during Poster Session 3b: Atrial Mapping, on 22 September from 12:30-14:00.

Thalia's paper, titled "Statistical Shape and Motion Model for Myocardial Deformation from Tagged Magnetic Resonance Imaging", analyses the motion of the heart muscle from tagged MRI. In the paper, she used tagged MRI to construct point clouds representing points in the myocardium and analysed this over a population derived from the UK Biobank to understand differences in myocardial deformation between individuals that are healthy and those with myocardial infarction, finding significant influence of changes to the twisting motion of the heart. Thalia presented this work during Poster Session 1: Cardiovascular Imaging, on 22 September from 17:15-18:45.

Adrian's paper, titled "End-to-End ECG Digitisation with Soft Segmentation and BiLSTM Signal Reconstruction", presents an end-to-end system for the digitisation of paper electrocardiographs (ECGs). Using a deep learning framework, photographs of paper ECGs are digitised by digitally removing creases, detecting the signals, and converting it into its time-series form. Building on top of the 2024 PhysioNet challenge and collaborating with clinicians, the digitiser was evaluated on real clinical ECG records and metrics, and was packaged into a mobile application. Adrian demonstrated this mobile app during his presentation, receiving enthusiastic feedback from several clinicians attending the conference. Adrian presented this work during Poster Session 4a: ECG segmentation, reconstruction, and QC, on 22 September from 17:15-18:45. Adrian also won the Gary and Bill Sanders Posters Award for his presentation and research.

We would like to congratulate all members of our lab who published work through CinC this year, and who gave presentations on their work during the conference. We would like to especially highlight and commend Adrian, who was given the Best Poster award for his work and accompanying poster presentation.