神马福利影片

The month of April saw the annual IEEE International Symposium on Biomedical Imaging (ISBI) 2026 held in London. This year, researchers and students had eight papers accepted Link opens in a new windowat the conference.

The first day of the conference, 9^th April, saw PhD student, Yijie Zhu, kick off the TIA Centre contributions. Yijie presented his poster entitled Joint Multiscale Image Learning for Segmentation in Macroscopic and Microscopic Pathology Images. Commenting on the experience 鈥淚 was happy to attend ISBI for the second time! It was a great experience to listen to many inspiring presentations and learn about cutting-edge methods in the field. I was also delighted that my paper was selected for a poster session, which gave me the opportunity to share my work with researchers interested in my topic in person and receive valuable feedback for my future work. It was also wonderful to meet new friends from all over the world鈥�.

At the same time undergraduate students Seth Chang (3rd year Undergraduate, Data Science) and Muhammad Amjad (4th year Undergraduate, Mathematics and Physics) had a fantastic opportunity to present a poster highlighting research they conducted whilst at the TIA Centre during an internship over the summer of 2025.Their poster was entitled Synergy vs. Noise: Performance-Guided Multimodal Fusion for Biochemical Recurrence-Free Survival in Prostate Cancer. A more detailed article about their experience at an international conference can be found hereLink opens in a new window.

Closing the first day鈥檚 activity was PhD student, Bashayer Abdallah who presented her paper entitled Monocular Depth Estimation with Guided Edge-Aware Attention for Endoscopic Images. Bashayer and colleagues developed a novel monocular depth estimation framework, DepthEdgeNet, designed to address one of the most persistent challenges in endoscopic imaging: preserving sharp, accurate depth boundaries around tissue edges.

鈥淪tandard depth estimation models tend to produce overly smooth predictions that blur critical structural discontinuities, which can compromise diagnostic and surgical precision. To address this, we proposed a lightweight end-to-end framework built around three key contributions: EdgeAttenLoss, an edge-aware attention loss that uses Scharr-derived gradient maps to re-weight supervision toward depth discontinuities during training; a Guidance Map Extractor that encodes RGB inputs into spatially rich structural cues; and a Guided-Affine Decoder that fuses encoder features with the learned guidance through per-pixel affine modulation, producing sharp, structure-consistent depth maps. Evaluated on the public UCL and C3VD colonoscopy datasets, DepthEdgeNet achieves state-of-the-art performance while running at a real-time inference speed of 17.32 ms per image, outpacing competing methods without sacrificing accuracy.鈥�

Bashayer also remarked that 鈥減resenting this work at IEEE ISBI 2026 was a truly memorable experience, an inspiring symposium at the forefront of biomedical imaging research, filled with thought-provoking discussions and a wonderful opportunity to connect with brilliant minds across the medical AI and computational imaging communities鈥�.

The second day of the conference started early for researchers Mark Eastwood, Research Fellow and Adam Shephard, Assistant Professor, who were both scheduled in the 8.00 am slot! Mark presented his research paper entitled Beer-Lambert Autoencoder for Unsupervised Stain Representation Learning and Deconvolution in Multi-Immunohistochemical Brightfield Histology Images and Adam presented on behalf of Hafsa Akebli, Visiting PhD Student from the University of Udine. Hafsa鈥檚 paper was entitled Multimodal Oncology Agent for IDH1 Mutation Prediction in Low-Grade Glioma and was included in the Special Session: Digital Twins and Multi-Omics Integration: Methodological Advances for Personalized Biomedical Modeling. Hafsa was delighted that she was awarded the Best Paper award at the conference 鈥� you can read more about her research hereLink opens in a new window and next steps.

A more leisurely start for the second day of the conference was had by one of our Research Fellows, Behnaz Elhaminia. Her poster was entitled A Deep Learning Framework for Glomeruli Instance Segmentation with Boundary Attention which focussed on AI-driven histopathology image segmentation and received a great deal of engagement during the session, with many people asking questions and showing interest in her work.

Behnaz shared 鈥渁ttending the IEEE ISBI 2026 in London was an exciting experience for me and gave me the opportunity to connect with researchers working on problems similar to my own research, which opened discussions around potential future collaborations and access to new datasets. Overall, it was a valuable experience that provided useful feedback, networking opportunities, and insight into the latest developments.鈥�

Wrapping up the second day for the TIA Centre was a PhD student, Kesi Xu, who presented his paper entitled Tissue Aware Nuclei Detection and Classification Model for Histopathology Images.

The final day of the ISBI Conference saw Research Fellow, Noorul Wahab, present his paper entitled ModalSurv: Investigating Opportunities and Limitations of Multimodal Deep Survival Learning in Prostate Cancer.

Overall, it was a fantastic few days for colleagues from the TIA Centre and it was a great opportunity to showcase the extent of the research being undertaken by students and Research Fellows at the Conference. We are already looking forward to next year鈥檚 conference!