Combining Shape and Learning for Medical Image Analysis
Doctoral thesis, 2020

Automatic methods with the ability to make accurate, fast and robust assessments of medical images are highly requested in medical research and clinical care. Excellent automatic algorithms are characterized by speed, allowing for scalability, and an accuracy comparable to an expert radiologist. They should produce morphologically and physiologically plausible results while generalizing well to unseen and rare anatomies. Still, there are few, if any, applications where today's automatic methods succeed to meet these requirements. 

The focus of this thesis is two tasks essential for enabling automatic medical image assessment, medical image segmentation and medical image registration. Medical image registration, i.e. aligning two separate medical images, is used as an important sub-routine in many image analysis tools as well as in image fusion, disease progress tracking and population statistics. Medical image segmentation, i.e. delineating anatomically or physiologically meaningful boundaries, is used for both diagnostic and visualization purposes in a wide range of applications, e.g. in computer-aided diagnosis and surgery.

The thesis comprises five papers addressing medical image registration and/or segmentation for a diverse set of applications and modalities, i.e. pericardium segmentation in cardiac CTA, brain region parcellation in MRI, multi-organ segmentation in CT, heart ventricle segmentation in cardiac ultrasound and tau PET registration. The five papers propose competitive registration and segmentation methods enabled by machine learning techniques, e.g. random decision forests and convolutional neural networks, as well as by shape modelling, e.g. multi-atlas segmentation and conditional random fields.

feature-based registration

convolutional neural networks

conditional random fields

medical image segmentation

random decision forests

machine learning

multi-atlas segmentation

medical image registration

shape models

Room EA, Hörsalsvägen 11
Opponent: Professor Marleen de Bruijne, University of Copenhagen, Denmark & Erasmus University Medical Center, the Netherlands


Jennifer Alvén

Chalmers, Electrical Engineering, Signal Processing and Biomedical Engineering

A Deep Learning Approach to MR-less Spatial Normalization for Tau PET Images

Medical Image Computing and Computer Assisted Intervention – MICCAI 2019. MICCAI 2019. Lecture Notes in Computer Science,; Vol. 11765(2019)p. 355-363

Paper in proceeding

Shape-aware label fusion for multi-atlas frameworks

Pattern Recognition Letters,; Vol. 124(2019)p. 109-117

Journal article

Max-margin learning of deep structured models for semantic segmentation

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics),; Vol. 10270 LNCS(2017)p. 28-40

Paper in proceeding

Automatic pericardium segmentation and quantification of epicardial fat from computed tomography angiography

Journal of Medical Imaging,; Vol. 3(2016)p. Article number 034003-

Journal article

Überatlas: Fast and robust registration for multi-atlas segmentation

Pattern Recognition Letters,; Vol. 80(2016)p. 249-255

Journal article

Medical imaging (x-ray, ultrasound, ...) allows scientists and physicians to examine, diagnose and treat diseases. The amount of medical images acquired on a daily basis is steadily growing. This trend will most definitely continue, due to an aging population, and an increased access to technological healthcare. The field of medical image analysis can help to reduce the manual workload by developing automatic algorithms for medical image interpretation. Such automatic assessments are useful for both medical research and for clinical applications, for example in computer-aided diagnosis, treatment planning and computer-assisted surgery. 

This thesis focuses on two fundamental tasks in the field of medical image analysis, medical image segmentation and medical image registration. Both tasks require algorithms able to recognise all kinds of anatomy and physiology, both common and rare. To construct such algorithms, the thesis uses tools such as machine learning (a branch of AI) and shape modelling. The algorithms are evaluated on several different applications. One application is to estimate the risk of cardiovascular diseases such as heart attacks, by measuring the amount of fat inside the heart sac. Another application is to diagnose Alzheimer’s disease, by measuring the amount of abnormal ‘tau’ proteins in the brain. In a short term perspective, the developed algorithms could facilitate large-scale medical studies, and in the longer run, also clinical care.

Areas of Advance

Information and Communication Technology

Life Science Engineering (2010-2018)

Subject Categories

Computer Vision and Robotics (Autonomous Systems)

Medical Image Processing



Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 4701



Room EA, Hörsalsvägen 11

Opponent: Professor Marleen de Bruijne, University of Copenhagen, Denmark & Erasmus University Medical Center, the Netherlands

More information

Latest update