Imperial College London


Faculty of MedicineInstitute of Clinical Sciences

Reader in Imaging Sciences



+44 (0)20 3313 1510declan.oregan




Imaging Sciences DepartmentHammersmith HospitalHammersmith Campus





Publication Type

124 results found

O'Regan DP, Fitzgerald J, Allsop J, Gibson D, Larkman DJ, Cokkinos D, Hajnal JV, Schmitz SAet al., 2005, A comparison of MR cholangiopancreatography at 1.5 and 3.0 Tesla, BRITISH JOURNAL OF RADIOLOGY, Vol: 78, Pages: 894-898, ISSN: 0007-1285

Journal article

Blunt D, O'Regan D, 2005, Using PACS as a teaching resource, BRITISH JOURNAL OF RADIOLOGY, Vol: 78, Pages: 483-484, ISSN: 0007-1285

Journal article

O'Regan D, Tait P, 2005, Imaging of the jaundiced patient, HOSPITAL MEDICINE, Vol: 66, Pages: 17-22, ISSN: 1462-3935

Journal article

Biffi C, Cerrolaza JJ, Tarroni G, Bai W, Oktay O, Folgoc LL, Kamnitsas K, Marvao AD, Doumou G, Duan J, Prasad SK, Cook SA, O'Regan DP, Rueckert Det al., Explainable Shape Analysis through Deep Hierarchical Generative Models: Application to Cardiac Remodeling

Quantification of anatomical shape changes still relies on scalar globalindexes which are largely insensitive to regional or asymmetric modifications.Accurate assessment of pathology-driven anatomical remodeling is a crucial stepfor the diagnosis and treatment of heart conditions. Deep learning approacheshave recently achieved wide success in the analysis of medical images, but theylack interpretability in the feature extraction and decision processes. In thiswork, we propose a new interpretable deep learning model for shape analysis. Inparticular, we exploit deep generative networks to model a population ofanatomical segmentations through a hierarchy of conditional latent variables.At the highest level of this hierarchy, a two-dimensional latent space issimultaneously optimised to discriminate distinct clinical conditions, enablingthe direct visualisation of the classification space. Moreover, the anatomicalvariability encoded by this discriminative latent space can be visualised inthe segmentation space thanks to the generative properties of the model, makingthe classification task transparent. This approach yielded high accuracy in thecategorisation of healthy and remodelled hearts when tested on unseensegmentations from our own multi-centre dataset as well as in an externalvalidation set. More importantly, it enabled the visualisation inthree-dimensions of the most discriminative anatomical features between the twoconditions. The proposed approach scales effectively to large populations,facilitating high-throughput analysis of normal anatomy and pathology inlarge-scale studies of volumetric imaging.

Working paper

This data is extracted from the Web of Science and reproduced under a licence from Thomson Reuters. You may not copy or re-distribute this data in whole or in part without the written consent of the Science business of Thomson Reuters.

Request URL: Request URI: /respub/WEB-INF/jsp/search-html.jsp Query String: id=00429325&limit=30&person=true&page=5&respub-action=search.html