Development and application of quantitative image analysis for preclinical MRI research

Al-Mubarak, Haitham (2020) Development and application of quantitative image analysis for preclinical MRI research. PhD thesis, University of Glasgow.

Abstract

The aim of this thesis is to develop quantitative analysis methods to validate MRI and improve the detection of tumour infiltration. The major components include a description of the development the quantitative methods to better validate imaging biomarkers and detect of infiltration of tumour cells into normal tissue, which were then applied to a mouse model of glioblastoma invasion. To do this, a new histology model, called Stacked In-plane Histology (SIH), was developed to allow a quantitative analysis of MRI.
Validating imaging biomarkers for glioblastoma infiltration
Cancer can be defined as a disease in which a group of abnormal cells grow uncontrollably, often with fatal outcomes. According to (Cancer research UK, 2019), there are more than 363,000 new cancer cases in the UK every year, an increase from the 990 cases reported daily in 2014-2016, with only half of all patients recovering.
Glioblastoma (GB) is the most frequent and malignant form of primary brain tumours with a very poor prognosis. Even with the development of modern diagnostic strategies and new therapies, the five-year survival rate is just 5%, with the median survival time only 14 months.
Unfortunately, glioblastoma can affect patients at any age, including young children, but has a peak occurrence between the ages of 65 and 75 years. The standard treatment for GB consists of surgical resection, followed by radiotherapy and chemotherapy. However, the infiltration of GB cells into healthy adjacent brain tissue is a major obstacle for successful treatment, making complete removal of a tumour by surgery a difficult task, with the potential for tumour recurrence.
Magnetic Resonance Imaging (MRI) is a non-invasive, multipurpose imaging tool used for the diagnosis and monitoring of cancerous tumours. It can provide morphological, physiological, and metabolic information about the tumour. Currently, MRI is the standard diagnostic tool for GB before the pathological examination of tissue from surgical resection or biopsy specimens. The standard MRI sequences used for diagnosis of GB include T2-Weighted (T2W), T1-Weighted (T1W), Fluid-Attenuated Inversion Recovery (FLAIR), and Contrast Enhance T1 gadolinium (CE-T1) scans. These conventional scans are used to localize the tumour, in addition to associated oedema and necrosis. Although these scans can identify the bulk of the tumour, it is known that they do not detect regions infiltrated by GB cells.
The MRI signal depends upon many physical parameters including water content, local structure, tumbling rates, diffusion, and hypoxia (Dominietto, 2014). There has been considerable interest in identifying whether such biologically indirect image contrasts can be used as non-invasive imaging biomarkers, either for normal biological functions, pathogenic processes or pharmacological responses to therapeutic interventions (Atkinson et al., 2001). In fact, when new MRI methods are proposed as imaging biomarkers of particular diseases, it is crucial that they are validated against histopathology. In humans, such validation is limited to a biopsy, which is the gold standard of diagnosis for most types of cancer. Some types of biopsies can take an image-guided approach using MRI, Computed Tomography (CT) or Ultrasound (US). However, a biopsy may miss the most malignant region of the tumour and is difficult to repeat. Biomarker validation can be performed in preclinical disease models, where the animal can be terminated immediately after imaging for histological analysis. Here, in principle, co-registration of the biomarker images with the histopathology would allow for direct validation. However, in practice, most preclinical validation studies have been limited to using simple visual comparisons to assess the correlation between the imaging biomarker and underlying histopathology.
First objective (Chapter 5): Histopathology is the gold standard for assessing non-invasive imaging biomarkers, with most validation approaches involving a qualitative visual inspection. To allow a more quantitative analysis, previous studies have attempted to co-register MRI with histology. However, these studies have focused on developing better algorithms to deal with the distortions common in histology sections. By contrast, we have taken an approach to improve the quality of the histological processing and analysis, for example, by taking into account the imaging slice orientation and thickness. Multiple histology sections were cut in the MR imaging plane to produce a Stacked In-plane Histology (SIH) map. This approach, which is applied to the next two objectives, creates a histopathology map which can be used as the gold standard to quantitatively validate imaging biomarkers.
Second objective (Chapter 6): Glioblastoma is the most malignant form of primary brain tumour and recurrence following treatment is common. Non-invasive MR imaging is an important component of brain tumour diagnosis and treatment planning. Unfortunately, clinic MRI (T1W, T2W, CE-T1, and FLAIR) fails to detect regions of glioblastoma cell infiltration beyond the solid tumour region identified by contrast enhanced T1 scans. However, advanced MRI techniques such as Arterial Spin Labelling (ASL) could provide us with extra information (perfusion) which may allow better detection of infiltration. In order to assess whether local perfusion perturbation could provide a useful biomarker for glioblastoma cell infiltration, we quantitatively analysed the correlation between perfusion MRI (ASL) and stacked in-plane histology. This work used a mouse model of glioblastoma that mimics the infiltrative behaviour found in human patients. The results demonstrate the ability of perfusion imaging to probe regions of low tumour cell infiltration, while confirming the sensitivity limitations of clinic imaging modalities.
Third objective (Chapter 7): It is widely hypothesised that Multiparametric MRI (mpMRI), can extract more information than is obtained from the constituent individual MR images, by reconstructing a single map that contains complementary information. Using the MRI and histology dataset from objective 2, we used a multi-regression algorithm to reconstruct a single map which was highly correlated (r>0.6) with histology. The results are promising, showing that mpMRI can better predict the whole tumour region, including the region of tumour cell infiltration.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Administrative and financial support from The Ministry of Higher Education and Scientific Research in Iraq.
• Keywords: MRI, glioblastoma, multi-parametric MRI, histology, tumour infiltration, stacked in-plane histology.
• Subjects: R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer)
• Colleges/Schools: College of Medical Veterinary and Life Sciences > School of Psychology & Neuroscience
• Supervisor's Name: Holmes, Dr. William and Foster, Dr. John
• Date of Award: 2020
• Depositing User: Mr Haitham Al-mubarak
• Unique ID: glathesis:2020-81370
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 18 May 2020 15:26
• Last Modified: 29 Aug 2022 08:26
• Thesis DOI: 10.5525/gla.thesis.81370
• URI: https://theses.gla.ac.uk/id/eprint/81370

Related URLs

• Article DOI
• Article DOI
• Article DOI