Pemberton, Hugh; (2022) Development and validation of quantitative MRI reports to support neuroradiological diagnosis of neurological disorders. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text HughPemberton-710194-PhD_Thesis.pdf - Accepted Version Download (38MB) | Preview

Abstract

Magnetic resonance imaging (MRI) has been used for decades to diagnose and track numerous diseases. It is the cornerstone of diagnosis in several aspects of clinical neuroradiology. In research settings, quantification of various structural, pathological and functional MRI features is used ubiquitously. However, this quantification is rarely used in purely clinical settings. Visual assessment and structured ratings scales are generally used instead, although clinical translation of quantitative methods is on the rise. With the workload of radiologists also increasing, automated quantification and well-established research technology are greatly needed in the clinic. In addition, contextualising single-patient findings to normative data offers enhanced objectivity and consistency in the diagnosis and treatment monitoring of neurological disorders. In this thesis, quantitative neuroradiological MRI reports (QReports) at various stages of development and validation are described and assessed for application in dementia, epilepsy (hippocampal sclerosis) and brain tumours (gliomas). The broad purpose of these QReports is to assist neuroradiological assessment and diagnosis. QReports display individual patient data, for example the degree of atrophy in a specific brain region, in the context of a normative population. Translation into clinical settings should follow a structured framework of development, including technical and clinical validation steps proceeded by in-use evaluation and health economics. The Quantitative Neuroradiology Initiative (QNI) framework, developed as part of this thesis, highlights six necessary steps for the development, validation and integration of QReports in clinic. While several such tools are currently available for dementia MRI assessment, this PhD’s systematic review highlights a significant evidence gap regarding clinical validation in the literature. The ensuing chapters aim to start plugging that gap by testing how QReports affect diagnostic accuracy and confidence in image raters of varying experience levels for dementia and epilepsy. For gliomas, the segmentation accuracy of three automated deep learning algorithms has been compared in order to guide the early-stage development of a glioma-centric QReport. The overall goal of these studies was to evaluate the technical and clinical validity of QReports through rigorous multi-centre studies. The broad hypotheses of this PhD were that quantification of clinically relevant data contextualised with age-appropriate normative populations will increase the precision of cross-sectional assessments by providing global and sub-regional information in addition to visual assessment, and generate more consistent reporting across different levels of rater experience. In summary, this work has found that using automated QReports alongside routine visual assessment provides significant improvements in diagnostic sensitivity, accuracy, confidence and inter-rater agreement for detecting abnormality in dementia and epilepsy cohorts. Several caveats exist and are discussed in detail. They highlights the need for rigorous validation on real world data before clinical adoption, careful monitoring of QReport translation and the importance of appropriate user training to avoid over-reliance on these diagnostic aids. Furthermore, the thesis expands upon current knowledge of how QReports affect neuroradiological MRI assessment and provides valuable information regarding the QReports currently available for diagnosis of neurological disorders.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item