Evaluating the impact of segmentation strategies on radiomic feature stability for paediatric brain tumour diagnosis using diffusion weighted imaging
Timothy Mulvany, Daniel Griffiths-King, Lara Worthington, Katherine Crombie, Heather E L Rose, Andrew Peet, John Apps, Jan Novak · 2026
Read Evaluating the impact of segmentation strategies on radiomic feature stability for paediatric brain tumour diagnosis using diffusion weighted imaging by Timothy Mulvany, Daniel Griffiths-King, Lara Worthington, Katherine Crombie, Heather E L Rose, Andrew Peet, John Apps, Jan Novak (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0356382.txt.
Description
Timothy Mulvany and seven colleagues test how the way a tumor is outlined on a scan affects computerized measurements used to diagnose children's brain tumors, in this 2026 study. Radiomics extracts numbers from MRI scans that machine learning can turn into diagnoses, but humans draw tumor boundaries differently. Using scans from just over one hundred young patients, the team shrank and expanded expert-drawn outlines to imitate cautious and generous drawing, then watched what happened to the measurements.
Expanding outlines disturbed nearly all measurements, and far more than shrinking them did. Diagnosis models trained on expert outlines lost accuracy under either change, but less when training included the altered outlines and only the steadiest measurements. The authors advise that when boundaries are ambiguous, leaving doubtful tissue out beats including it, and that model builders should plan for drawing variation from the start.
Catalog record
| CREATOR | Timothy Mulvany, Daniel Griffiths-King, Lara Worthington, Katherine Crombie, Heather E L Rose, Andrew Peet, John Apps, Jan Novak |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31200 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |