Methods for quantitative analysis of myocardial perfusion SPECT : validated with magnetic resonance imaging, phantom studies and expert readers

Research output: ThesisDoctoral Thesis (compilation)

Abstract

In this thesis, methods for automated analysis of myocardial perfusion single photon emission computer tomography (myocardial perfusion SPECT, MPS) images have been successfully developed and evaluated. MPS images show the perfusion of the myocardium and are used for diagnosing patients with suspected ischemic heart disease. The disease is characterized by reduced perfusion to a region of the myocardium, showed as perfusion defects in MPS images. The reduced perfusion causes a dysfunctional cardiac pumping, and can result in cardiac cell death. The disease is a major cause of death and disability worldwide. However, by an early diagnosis and correct patient management, the condition is treatable. Diagnosing and evaluation of treatments for ischemic heart disease is therefore of great importance. The aim of this thesis was to provide methods for automatic image analysis of MPS images, with high agreement to reference standards and adaptive for different research purposes.

The challenges of a method for automated analysis of MPS are the low image resolution, noisy image data, influence from other tissues in the image, and information loss in the image due to perfusion defects. To handle these difficulties in the analysis process, the proposed method complements the intensity information in the images with mathematical methods and knowledge of cardiac anatomy.
The mathematical methods helps to find edges in noisy image data, filter out irrelevant image information, smooth surfaces that separate different tissues in the image, and interpolates to overcome information loss. The anatomic knowledge adds constraints to ensure that the estimated volumes of the heart are physiologically reasonable. Furthermore, by training and optimizing the algorithms, the computer learns to identify the heart in the image, and the patterns for a myocardial
region with normal and reduced perfusion, respectively.

The proposed method for analysis of MPS images have been evaluated in a large number of patients. The results show that the proposed method can be used to quantify left ventricular volumes with low bias compared to the reference standard, cardiac magnetic resonance (CMR). Furthermore, the bias for left ventricular volumes was significantly lower by the proposed method compared to four other commercially available MPS software packages. It was also found that acute perfusion defects, called myocardium at risk (MaR), quantified by the proposed method show high agreement to manual delineation of MaR in MPS, which is considered the present reference standard. Moreover, the bias for quantification of MaR compared to the reference was lower by the proposed method than for a commercially available MPS software package. The results illustrate the benefit of the proposed method specifically developed for segmentation of MaR, compared to the other method developed for identification of varying degree of ischemia. Finally, the proposed method could be trained to detect and quantify stress-induced and chronic perfusion defects, called stress-induced ischemia and myocardial infarction, respectively, with good correlation to expert readers and true values from simulated MPS data. In conclusion, the proposed methods can be used to asses left ventricular volumes, myocardium at risk, and stress-induced ischemia as well as infarction from MPS images.
Original languageEnglish
QualificationDoctor
Awarding Institution
  • Mathematics (Faculty of Engineering)
Supervisors/Advisors
  • Heiberg, Einar, Supervisor
Award date2012 Mar 30
Publisher
Print ISBNs978-91-7473-285-6
Publication statusPublished - 2012

Bibliographical note

Defence details

Date: 2012-03-30
Time: 09:15
Place: MH:B Matematikcentrum

External reviewer(s)

Name: Underwood, S. Richard
Title: Professor
Affiliation: Imperial College London, London, UK

---




The information about affiliations in this record was updated in December 2015.
The record was previously connected to the following departments: Numerical Analysis (011015004)

Subject classification (UKÄ)

  • Mathematics

Keywords

  • Segmentation
  • Quantification
  • Spline
  • Image registration
  • Cardiac image analysis
  • Myocardial perfusion SPECT

Fingerprint

Dive into the research topics of 'Methods for quantitative analysis of myocardial perfusion SPECT : validated with magnetic resonance imaging, phantom studies and expert readers'. Together they form a unique fingerprint.

Cite this