Applying machine learning classifiers to automate quality assessment of paediatric dynamic susceptibility contrast (DSC-) MRI data - Full paper

Aug 3, 2026

Upload by Logic

Dynamic susceptibility contrast (DSC-) MRI enables perfusion estimation in paediatric brains but reliable quality assessment is typically performed through qualitative review (QR). This study evaluates QR performance and develops an automated alternative using machine learning classifiers trained on signal–time course quality metrics derived from 1027 courses. Inter-reviewer agreement (κ=0.83) supports QR consistency, while thresholded measures and a random forest model demonstrate strong sensitivity, specificity, precision, classification error, and AUC. Combining multiple measures reduces misclassification.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Objective

Methods

Data collection and qualitative review

Signal–time course measures and thresholds

Classifier training and evaluation

Results

Conclusion

Advances in knowledge

BJR [https://doi.org/10.1259/bjr.20201465](https://doi.org/10.1259/bjr.20201465) |
Received:
21 December 2020
Accepted:
24 January 2023
Published online:
16 February 2023
© 2023 The Authors. Published by the British Institute of Radiology under the terms of the Creative Commons Attribution 4.0 Unported License [http://creativecommons.org/l](http://creativecommons.org/l)icenses/by/4 .0/, which permits unrestricted use, distribution and reproduction in any medium, provided the original author and source are credited.
Cite this article as:
Powell SJ, Withey SB, Sun Y, Grist JT, Novak J, MacPherson L, et al. Applying machine learning classifiers to automate quality assessment of paediatric dynamic susceptibility contrast (DSC-) MRI data. Br J Radiol (2023) 10.1259/bjr.20201465. | | | --- | --- | | FULL PAPER
Applying machine learning classifiers to automate quality assessment of paediatric dynamic susceptibility contrast (DSC-) MRI data
1,2STEPHEN J. POWELL, 2,3,4STEPHANIE B. WITHEY, 2,5YU SUN, 2JAMES T. GRIST, 2,3,6JAN NOVAK,
7LESLEY MACPHERSON, 8LAURENCE ABERNETHY, 9BARRY PIZER, 10RICHARD GRUNDY, 10,11,12PAUL S. MORGAN, 10,13TIM JASPAN, 14SIMON BAILEY, 15DIPAYAN MITRA, 16DOROTHEE P. AUER, 8SHIVARAM AVULA,
2,3,17THEODOROS N. ARVANITIS and 2,3ANDREW PEET
1Physical Sciences for Health CDT, University of Birmingham, Birmingham, United Kingdom
2 Institute of Cancer and Genomic Sciences, University of Birmingham, Birmingham, United Kingdom
3Department of Oncology, Birmingham Children’s Hospital, Birmingham, United Kingdom
4RRPPS, University Hospitals Birmingham NHS Foundation Trust, Birmingham, United Kingdom
5School of Biological Sciences and Medical Engineering, Southeast University, Nanjing, China
6Department of Psychology, Aston Brain Centre, School of Life and Health Sciences, Aston University, Birmingham, United Kingdom
7Radiology, Birmingham Children’s Hospital, Birmingham, United Kingdom
8Radiology, Alder Hey Children’s NHS Foundation Trust, Liverpool, United Kingdom
9Oncology, Alder Hey Children’s NHS Foundation Trust, Liverpool, United Kingdom
10The Children’s Brain Tumour Research Centre, University of Nottingham, Nottingham, United Kingdom 11Medical Physics, Nottingham University Hospitals, Nottingham, United Kingdom
12NIHR Nottingham Biomedical Research Centre, Nottingham, United Kingdom
13Radiology, Nottingham University Hospitals, Nottingham, United Kingdom
14Sir James Spence Institute of Child Health, Royal Victoria Infirmary, Newcastle upon Tyne, United Kingdom
15Neuroradiology, The Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, United Kingdom
16Sir Peter Mansfield Imaging Centre, University of Nottingham, Nottingham, United Kingdom
17 Institute of Digital Healthcare, WMG, University of Warwick, Coventry, United Kingdom
Address correspondence to: Professor Andrew Peet
E-mail: [a.peet@bham.ac.uk](a.peet@bham.ac.uk) | | | Objective: Investigate the performance of qualitative review (QR) for assessing dynamic susceptibility contrast (DSC-) MRI data quality in paediatric normal brain and develop an automated alternative to QR. Methods: 1027 signal–time courses were assessed by Reviewer 1 using QR. 243 were additionally assessed by Reviewer 2 and % disagreements and Cohen’s κ (κ) were calculated. The signal drop-to-noise ratio (SDNR), root mean square error (RMSE), full width half maximum (FWHM) and percentage signal recovery (PSR) were calculated for the 1027 signal–time courses. Data quality thresholds for each measure were determined using QR results. The measures and QR results trained machine learning classifiers. Sensitivity, specificity, precision, classification error and area under the curve from a receiver operating characteristic curve were calculated for each threshold and classifier.
Results: Comparing reviewers gave 7% disagreements and κ = 0.83. Data quality thresholds of: 7.6 for S

FAQs

What problem does the study address in paediatric DSC-MRI quality assessment?

It evaluates how well qualitative review (QR) assesses DSC-MRI data quality in paediatric normal brain and develops an automated alternative to replace or augment QR.

How were quality thresholds and machine learning classifiers built?

Quality thresholds for metrics such as SDNR, RMSE, FWHM, and PSR were derived from QR results, and the metrics together with QR outcomes were used to train machine learning classifiers.

Which classifier and measure performed best, and what was the key outcome?

Random forest was the best classifier, while SDNR produced the best overall metric performance. The study concludes that machine learning classifiers trained on signal–time course measures and QR can assess quality and that combining measures reduces misclassification.

Aug 3, 2026

Upload by Logic