Automatic Pancreas Segmentation and 3D Reconstruction for Morphological Feature Extraction in Medical Image Analysis

Asaturyan, H. 2021. Automatic Pancreas Segmentation and 3D Reconstruction for Morphological Feature Extraction in Medical Image Analysis. PhD thesis University of Westminster Computer Science and Engineering

TitleAutomatic Pancreas Segmentation and 3D Reconstruction for Morphological Feature Extraction in Medical Image Analysis
TypePhD thesis
AuthorsAsaturyan, H.
Abstract

The development of highly accurate, quantitative automatic medical image segmentation techniques, in comparison to manual techniques, remains a constant challenge for medical image analysis. In particular, segmenting the pancreas from an abdominal scan presents additional difficulties: this particular organ has very high anatomical variability, and a full inspection is problematic due to the location of the pancreas behind the stomach. Therefore, accurate, automatic pancreas segmentation can consequently yield quantitative morphological measures such as volume and curvature, supporting biomedical research to establish the severity and progression of a condition, such as type 2 diabetes mellitus. Furthermore, it can also guide subject stratification after diagnosis or before clinical trials, and help shed additional light on detecting early signs of pancreatic cancer. This PhD thesis delivers a novel approach for automatic, accurate quantitative pancreas segmentation in mostly but not exclusively Magnetic Resonance Imaging (MRI), by harnessing the advantages of machine learning and classical image processing in computer vision. The proposed approach is evaluated on two MRI datasets containing 216 and 132 image volumes, achieving a mean Dice similarity coefficient (DSC) of 84:1  4:6% and 85:7  2:3% respectively. In order to demonstrate the universality of the approach, a dataset containing 82 Computer Tomography (CT) image volumes is also evaluated and achieves mean DSC of 83:1  5:3%. The proposed approach delivers a contribution to computer science (computer vision) in medical image analysis, reporting better quantitative pancreas segmentation results in comparison to other state-of-the-art techniques, and also captures detailed pancreas boundaries as verified by two independent experts in radiology and radiography. The contributions’ impact can support the usage of computational methods in biomedical research with a clinical translation; for example, the pancreas volume provides a prognostic biomarker about the severity of type 2 diabetes mellitus. Furthermore, a generalisation of the proposed segmentation approach successfully extends to other anatomical structures, including the kidneys, liver and iliopsoas muscles using different MRI sequences. Thus, the proposed approach can incorporate into the development of a computational tool to support radiological interpretations of MRI scans obtained using different sequences by providing a “second opinion”, help reduce possible misdiagnosis, and consequently, provide enhanced guidance towards targeted treatment planning.

Year2021
File
File Access Level
Open (open metadata and files)
Publication dates
Published01 Mar 2021

Related outputs

A Framework for Automatic Morphological Feature Extraction and Analysis of Abdominal Organs in MRI Volumes
Asaturyan, H., Thomas, E.L., Bell, J.D. and Villarini, B. 2019. A Framework for Automatic Morphological Feature Extraction and Analysis of Abdominal Organs in MRI Volumes. Journal of Medical Systems. 43 334. https://doi.org/10.1007/s10916-019-1474-3

Advancing Pancreas Segmentation in Multi-protocol MRI Volumes using Hausdorff-Sine Loss Function
Asaturyan, H., Thomas, E.L., Fitzpatrick, J., Bell, J.D. and Villarini, B. 2019. Advancing Pancreas Segmentation in Multi-protocol MRI Volumes using Hausdorff-Sine Loss Function. 10th International Workshop on Machine Learning in Medical Imaging (MLMI 2019) in conjunction with MICCAI 2019. Shenzen, China 13 Oct 2019 Springer. https://doi.org/10.1007/978-3-030-32692-0_4

Morphological and multi-level geometrical descriptor analysis in CT and MRI volumes for automatic pancreas segmentation
Asaturyan, H., Gligorievski, A. and Villarini, B. 2019. Morphological and multi-level geometrical descriptor analysis in CT and MRI volumes for automatic pancreas segmentation. Computerized Medical Imaging and Graphics. 75, pp. 1-13. https://doi.org/10.1016/j.compmedimag.2019.04.004

Hierarchical Framework for Automatic Pancreas Segmentation in MRI Using Continuous Max-flow and Min-Cuts Approach
Asaturyan, H. and Villarini, B. 2018. Hierarchical Framework for Automatic Pancreas Segmentation in MRI Using Continuous Max-flow and Min-Cuts Approach. ICIAR 2018 International Conference Image Analysis and Recognition. Póvoa de Varzim, Portugal 27 - 29 Jun 2018 Springer. https://doi.org/10.1007/978-3-319-93000-8_64

A Framework for Morphological Feature Extraction of Organs from MR Images for Detection and Classification of Abnormalities
Villarini, B., Asaturyan, H., Thomas, E.L., Mould, R. and Bell, J.D. 2017. A Framework for Morphological Feature Extraction of Organs from MR Images for Detection and Classification of Abnormalities. Proceedings of the 30th IEEE International Symposium on Computer-Based Medical Systems (CBMS’17). Thessaloniki, Greece 22 - 24 Jun 2017 IEEE . https://doi.org/10.1109/CBMS.2017.49

Permalink - https://westminsterresearch.westminster.ac.uk/item/v414y/automatic-pancreas-segmentation-and-3d-reconstruction-for-morphological-feature-extraction-in-medical-image-analysis


Share this
Tweet
Email

Usage statistics

21 total views
12 total downloads
2 views this month
2 downloads this month
These values are for the period from September 2nd 2018, when this repository was created

Export as