Pretreatment MRI texture features and discordance between early clinical and volumetric radiological response in primary central nervous system lymphoma
Pretreatment imaging biomarkers of early treatment response in primary central nervous system lymphoma (PCNSL) remain limited, and clinical improvement may not parallel radiological regression. This retrospective single-center study investigated whether pretreatment MRI texture features were associated with early radiological and clinical response in 46 patients with PCNSL treated with high-dose methotrexate-based induction therapy and characterized clinical-radiological discordance. Eleven prespecified first-order and gray-level co-occurrence matrix (GLCM) features were extracted from T2-weighted images, apparent diffusion coefficient maps, and contrast-enhanced T1-weighted images (CE-T1WI), with CE-T1WI GLCM homogeneity designated as the principal feature. Radiological response occurred in 21/46 patients (45.7%) and clinical response in 35/46 (76.1%). Fourteen patients improved clinically without meeting the volumetric radiological-response threshold, whereas the reverse pattern was not observed (κ = 0.42; McNemar p < .001). Higher pretreatment CE-T1WI GLCM homogeneity was associated with radiological response (OR per 0.10-unit increase, 3.41; 95% CI, 1.45–8.03; p = .005; AUC, 0.76) and remained significant after adjustment for baseline tumor volume (adjusted OR, 3.22; p = .010). Homogeneity was also associated with clinical response (OR, 3.86; p = .019; AUC, 0.75). Pretreatment CE-T1WI GLCM homogeneity may therefore represent a reproducible imaging feature associated with early response in PCNSL, while the frequent clinical-radiological discordance supports multidimensional assessment of early treatment response.
Authors
- Hakan Cebeci (ORCID: https://orcid.org/0000-0002-2017-3166)
- Halil Özer (ORCID: https://orcid.org/0000-0003-1141-1094)
- Cem Selim (ORCID: https://orcid.org/0000-0002-2717-8306)
- Ahmet Baytok (ORCID: https://orcid.org/0000-0003-1615-5771)
- Seyit Erol (ORCID: https://orcid.org/0000-0002-0547-872X)
- Ayşe Arı (ORCID: https://orcid.org/0000-0001-5778-115X)
- Elif Kamalak
Institutions
- Selçuk University (TR)
Publication Details
- Journal
- The Neuroradiology Journal
- Published
- 2026-11-21
- DOI
- https://doi.org/10.1177/19714009261509882
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00