Prize for best MRes Report 2015-16 awarded to Marta Ranzini
Congratulations to Marta Ranzini who has been awarded the Centre for Doctoral Training in Medical Imaging Prize for best MRes report 2015/2016 for her work entitled: “The use of 3D Imaging for Musculoskeletal Disease Computational Anatomy”
Abstract from the report:
Musculoskeletal conditions affect hundreds of millions of people all over the world. In the past 15 years, Metal-on-Metal hip atrhroplasty has been one of the most ef- fective surgical interventions in improving quality of life. However, this implant type is associated with high risk of failure, due to adverse inflammatory reactions and increased muscle atrophy. Routine assessment of periprosthetic muscles re- sponse to the implant is perfomed on MR images, whilst CT imaging is exploited for surgical planning and post-operative follow-up. Hence, a tool able to merge the complementary 3D information yielded by the two modalities could be greatly ben- eficial. The main aim of this work consisted in the development of a framework for au- tomated segmentation of bones, muscles and implants in combined CT and MR images of the same subject. Consequently, I aimed at assessing whether a patient- specific 3D reconstruction of the musculoskeletal anatomy of the pelvis and imaging biomarkers of muscular atrophy could be automatically extracted from the devel- oped methodology. To this purpose, I constructed two template data sets composed of semi-manually segmented and co-registered CT and MR images from 8 implanted and 6 non im- planted hip sides respectively. A fully automated pipeline was subsequently de- veloped to perfom multi-atlas based segmentation propagation and label fusion on unseen CT/MR images. This pipeline was optimized and validated in a leave-one out cross-validation study, showing an accuracy comparable with single-modality automated segmentation techniques. Finally, as a proof-of-concept of the frame- work clinical relevance, patient-specific 3D rendering of muscles and bones and a volumetric fat-to-muscle ratio have been obtained from the automated segmentation of 13 unseen subjects. Allowing high resolution visualisation of spatial localisation and inter-volume re- lationship of muscular and bony structures, the proposed tool shows promise for surgical planning customisation applications and prediction of clinical outcomes.