Upcoming MultiMeDIA Lab Presentations at MICCAI 2026

From 27th September to 1st October, the 27th International Conference on Medical Image Computing & Computer Assisted Intervention (MICCAI) will be taking place in Strasbourg, France. MICCAI is the premier international conference for research in computer aided medical image analysis, receiving a record number of submissions this year. The MultiMeDIA lab is excited to announce that two of our members, DPhil students Zhengda Ma and Deyu Meng have been selected for poster presentations at this year's conference.

Zhengda's paper, titled "HeartFormer: Semantic-Aware Dual-Structure Transformers for 3D Four-Chamber Cardiac Point Cloud Reconstruction" introduces HeartFormer, a transformer-based framework that reconstructs complete 3D four-chamber heart anatomy from routine cine MRI. Unlike existing methods that primarily focus on the ventricles, HeartFormer models both the global heart structure and individual cardiac substructures to generate anatomically consistent reconstructions from sparse and misaligned clinical images. Also presented is HeartCompv1, the first large-scale public benchmark for multi-class cardiac point cloud completion, containing 17,000 high-resolution 3D heart models. Extensive evaluation demonstrates state-of-the-art reconstruction accuracy and strong preservation of clinically relevant anatomy, advancing fully automated 3D cardiac modelling from standard clinical MRI. Zhengda will be presenting this work during Poster Session 2, on 29th September from 10:30-12:00.

Deyu's paper, titled "Multi-stage NeRF for Efficient 3D Coronary Artery Reconstruction from Two Narrow-Angle Angiographic Projections" introduces NeCA++, which reconstructs 3D coronary arteries from just two narrow-angle X-ray angiographic projections taken at clinical narrow angle settings. Its self-supervised, two-stage approach first identifies the likely vessel region and then progressively refines the vascular structure with vessel-specific constraints. A ray-aligned penalty reduces ambiguity along the X-ray paths, and a binary penalty accelerates the separation of blood vessels from the background. Across three datasets, NeCA++ outperformed comparison methods under clinical narrow-angle settings, and was able to complete reconstructions in less than one minute per case. Deyu will be presenting this work during Poster Session 2, on 29th September from 10:30-12:00.

We congratulate Zhengda and Deyu on these achievements, and wish them the best of luck with their upcoming poster presentations! The supporting papers to their presentations will be released in the upcoming MICCAI conference proceedings.