Integrated white-matter lesions and grey-matter shrinkage quantification for clinical use

Lead Research Organisation: University College London
Department Name: Institute of Neurology

Abstract

The aim of this project is to bridge the gap between advances in neuroimaging analysis and clinical applications in patients with multiple sclerosis (MS) and dementia with cerebrovascular damage. A collaboration between the UCL Institutes of Neurology and Healthcare Engineering (called quantitative neuroradiology initiative [QNI]) aims to translate the unique knowledge in imaging acquisition and analysis accumulated at UCL to normalise, segment and quantify MRI scans towards clinical applicability. Promising results have been obtained to quantify grey matter loss in Alzheimer's disease (AD) cross-sectionally. By using reference datasets, it is now possible to perform an automatic quality control of each patient's MRI scan, and quantify both brain volume (normalized to intracranial volume) and hippocampal volume. These parameters, which give an indication of the disease state and severity in individual patients, are automatically calculated and provided in a routine radiological report for clinical use by integrating it seamlessly into the NHS clinical picture archiving and communication system (PACS). While the dementia project is moving towards a clinical validation step, it is now time to expand the QNI concept to the cross-sectional and longitudinal assessment of MS and vascular dementia.
White matter lesions and brain atrophy are interrelated processes: while the presence of lesions impacts our ability to accurately estimate brain atrophy, morphological changes results in unstable longitudinal lesion measurements. Currently, algorithms solve either the atrophy estimation or the lesion segmentation problems separately using techniques based on discrete geometry and generative models respectively. This project will focus on the development of a fully integrated joint Bayesian model for white matter lesion segmentation, brain parcellation and atrophy estimation to produce unbiased and accurate biomarkers. The algorithm developed in this project will be deployed in a clinical setting, requiring their integration within the clinical workflow and an appropriate validation on relevant data through the establishment of a reliable reference dataset, such as the ADNI dataset for dementia.

Industry partnership and the EPSRC Strategy:
The QNI prototype developed for assessment of regional atrophy in dementia is being exploited by BrainMiner (http://brainminer.co.uk/), a spin-out company initiated by UCL for the use in memory clinics in the UK and abroad. The plan of this company is to expand their services in the future to other disease areas such as MS. Incorporation of WML in dementia is important for a comprehensive assessment of all cases of dementia, including vascular dementia. BrainMiner offers in-kind contributions for the development of CE-marked software, development of user requirements, integration into clinical environment, and will be happy to host the candidate for a yearly placement.
This project aligns well with the "Leading-edge healthcare and medicine" area of the EPSRC "Building Our Industrial Strategy" green paper by aiming to develop advanced image analysis techniques that will be integrated into the clinical environment, thus advancing the UK healthcare capabilities by improving the standard of care, while also providing better imaging biomarkers that can be used to more accurately measure therapeutic efficacy, impacting the development of novel medicines. Finally, this project also tackles important crosscutting ESPRC capabilities such as data science and Big Data analytics due to the nature and complexity of medical imaging and epidemiological data.

Publications

10 25 50
publication icon
Goodkin O (2019) The quantitative neuroradiology initiative framework: application to dementia. in The British journal of radiology

Studentship Projects

Project Reference Relationship Related To Start End Student Name
EP/R512400/1 01/10/2017 30/09/2021
1950257 Studentship EP/R512400/1 25/09/2017 30/09/2021 Hugh Pemberton
 
Description I have performed a proof of concept study and first steps towards clinical validation for a neuroradiological report highlights volume loss (grey-matter shrinkage quantification) with positive results and imminent publication - see abstract below (paper is under review at European Radiology). I will now include more radiologists and perform a large 'accuracy study' to further elucidate the clinical benefit of using the report. I am also now working on creating a longitudinal dementia neuroradiological report and the inclusion of white matter pathology.

Automated quantitative brain MRI volumetry reports support diagnostic interpretation in dementia: A multi-rater, proof-of-concept study:

Key points
- First clinical validation of novel quantitative brain MRI atrophy reports for use as a diagnostic aid in dementia
- Quantitative reports referencing single-subject results to normative data alongside visual assessment improve sensitivity for detecting abnormality
- Consultant neuroradiologists' assessment accuracy and kappa scores significantly improve with the use of quantitative atrophy reports

Abstract

Background:
Neuroradiological assessment for dementia relies on subjective visual interpretation of MR images. We examined whether providing a quantitative report of regional brain volumes improves accuracy and confidence in detecting volume loss, and in differentiating Alzheimer's Disease (AD) and Frontotemporal Dementia (FTD), compared with visual assessment alone.

Methods:
Our forced-choice proof-of-concept multi-rater study used MRI from: 16 AD patients, 14 FTD patients, and 15 healthy controls; age range 52-81. Quantitative reports were presented to raters with regional grey matter volumes plotted as percentiles against data from a normative population (n=461). Nine raters with varying radiological experience (3 each: consultants, registrars, 'non-clinical image analysts') assessed each case twice (with and without the report). Raters were blinded to clinical and demographic information; they classified scans as 'normal' or 'abnormal' and if 'abnormal' as 'AD' or 'FTD'.

Results:
The report improved sensitivity for detecting volume loss across all raters combined (p=0.015*). Only the consultant group's accuracy increased significantly when using the report (p=0.02*). Overall, raters' agreement (Cohen's ?) with the 'gold standard' was not significantly affected by the report; only the consultant group improved significantly (?s 0.41->0.55, p=0.04). Cronbach's alpha for inter-rater agreement improved from 0.886 to 0.925, corresponding to an improvement from 'good' to 'excellent'.

Conclusion:
Quantitative reports referencing single-subject results to normative data alongside visual assessment improve sensitivity, accuracy and inter-rater agreement for detecting volume loss. The quantitative report was most effective in the consultants, suggesting that experience is needed to fully benefit from the additional information provided by quantitative analyses.
Exploitation Route The reports that have been created are currently in use at the National Hospital for Neurology and Neurosurgery and will be used by other hospitals to assist diagnostics in dementia.
Sectors Healthcare

URL http://qni.cs.ucl.ac.uk/
 
Title Quantitiative Neuroradiological Report highlighting volume loss resulting from dementia 
Description Quantitiative Neuroradiological Report highlighting volume loss resulting from dementia that was testing on 9 individuals. Three groups of raters participated in this study: 1) consultant neuroradiologists; 2) neuroradiology specialty registrars); and 3) MRI radiographers and non-clinical image analysts (designated from now as "image analyst group"). Raters were invited from multiple centres, ensuring a broad representation of training and experience. Raters were blinded to all clinical and demographic information except age and gender. We designed a website platform (Figure 2) to facilitate remote participation. The website included thorough instructions followed by the 45 scans displayed in a randomly generated order, once with and once without the quantitative report (QReport)). The task thus consisted of 90 evaluation 'episodes' in total. At each 'episode', raters were prompted to give their assessment, stating 1.a) whether the scan was 'normal' or 'abnormal' in terms of volume loss for age; 1.b) their degree of confidence on a scale of 1 (very uncertain) - 5 (very confident); 2.a) if the scan was rated abnormal (volume loss), to select AD or FTD as the most likely diagnosis; and 2.b) their confidence level for this differential diagnosis (1-5 scale). Raters completed the exercise over a period of two months, their ratings were collected through the web platform and subsequently analysed. We are currently making some changes to the report and will the retest this in 40 neuroradiologists. 
Type Diagnostic Tool - Imaging
Current Stage Of Development Early clinical assessment
Year Development Stage Completed 2019
Development Status Under active development/distribution
Clinical Trial? Yes
Impact 2 research papers under review and 2 more in progress. 
URL http://qni.cs.ucl.ac.uk/
 
Description Dragon's Den Enterprise Case Studies 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Undergraduate students
Results and Impact In this scenario, UCL undergraduate Biomedical Engineering students work in teams of 4-6 to devise a conceptual commercial product or service, build a business plan, and pitch this business plan to a panel of "Dragons" at the end of the week.

Each team is assigned a specific research/technology area and a contact researcher within UCL, who they will meet ideally in the first few days of the week to gather information and ask questions about the research. They can then follow-up with questions, but these are limited in number.

I supervised a team of undergraduates and provided them with my quantitative neuroradiological report for their presentations - and they won the competition.
Year(s) Of Engagement Activity 2019