AI and Cancer Research ── Brain ── 2026-11-01
Brain tumours: treatment, trials and AI
- Abstract
Background: Therapy-induced tissue alterations complicate intraoperative diagnosis of recurrent CNS tumors. We evaluated stimulated Raman histology (SRH) combined with convolutional neural networks (CNNs) for rapid tumor detection and classification.
Methods: In this bicentric observational study, fresh tissue from patients undergoing resection or stereotactic biopsy for suspected recurrent CNS tumors was imaged with fiber-laser-based SRH. CNN1 performed binary tumor/non-tumor prediction and CNN2 histotype classification. CNN1 was evaluated against blinded neuropathologist SRH consensus and CNN2 against the final integrated histopathological diagnosis.
Results: Of 382 acquired SRH images, 262 images from 143 patients were included. CNN1 achieved an AUC of 0.975 (95% CI, 0.95–1.00) for tumor detection. An exploratory cohort-derived tumor probability threshold of ≥ 59 increased specificity from 70.7% to 92.0% while maintaining 95.2% sensitivity and identifying 71.4% of radiation necrosis samples. Among 212 images in treatment subgroup analysis, CNN1 accuracy did not differ significantly across prior-treatment groups (overall 90.1%; p = 0.080). CNN2 accuracy declined from 61.8% after surgery alone to 39.0% after surgery, radiotherapy, and medical treatment (χ²=6.04, p = 0.049). Multimodal therapy was independently associated with lower odds of correct CNN2 prediction (adjusted OR, 0.10 [95% CI, 0.02–0.39]; p = 0.001 and 0.04 [95% CI, 0.01–0.20]; p < 0.001). 120/382 images (31.5%) were of insufficient diagnostic quality; application of the quality control threshold would have rejected 84/120 (70.0%).
Conclusions: CNN1 maintained high tumor/non-tumor discrimination across treatment subgroups when image quality was sufficient, whereas CNN2 entity-classification performance declined with increasing treatment intensity. Reliable SRH-based diagnosis in recurrent tumors requires validated image-quality control measures and model adaptation to recurrent disease.
Journal IF-equivalent: 1.2 (OpenAlex 2-year mean citedness, value as of 2026-10-31, retrieved 2026-11-01; not the official Clarivate IF)Reference: Fürtjes G, Al Shughri A, Stehle T, Reinecke D, Meißner AK, Jung N, et al. Performance of intraoperative AI-assisted Label-Free Stimulated Raman Histology in Recurrent Brain Tumors. Neuro-Oncology Advances. 2026 Oct 29 [Epub ahead of print]. doi:10.1093/noajnl/vdag267.Checked: Abstract only - Abstract
Background: Distinguishing true progression (TP) from pseudoprogression (PsP) in glioblastoma (GBM) remains a clinical challenge as both often manifest as indeterminate contrast-enhancing lesions on post-treatment MRIs. However, integrating dynamic contrast-enhanced (DCE)-MRI with analytical methods such as clustering and machine learning offers a promising approach for distinguishing TP from PsP.
Methods: Quantitative analysis of DCE-MRI data from 82 patients with TP (n = 53) and PsP (n = 29) was performed using five pharmacokinetic models. Subsequently, a parsimonious selection method was used to identify the best fit for each voxel. In next step, clustering analysis using K-means ++ was performed to classify pixels into low, medium, and high-intensity regions. Tumors were segmented into enhancing, non-enhancing, and edema regions using automated nnU-Net algorithm and clusters were extracted from these regions. Significant features derived from clustering analysis and descriptive statistics (mean, median, and skewness) of DCE-MRI parameters and tumor volume were incorporated as input variables for classifier development. Finally, five machine learning classifiers were employed to differentiate between TP and PsP.
Results: Univariable clustering analysis of DCE-MRI data identified Ktrans, Ve, Vp, τi and tumor volumes as significant parameters for distinguishing TP from PsP. Multivariable clustering revealed that combined high-Ktrans–high-Ve is a key distinctive variable. Among machine learning classifiers tested using these parameters, Random Forest demonstrated the best diagnostic performance, achieving an accuracy of 83% and an AUC of 0.85.
Conclusion: Parsimonious analysis of DCE-MRI data in conjunction with clustering and machine learning could serve as a valuable method for accurately distinguishing TP from PsP in GBMs.
Journal IF-equivalent: 1.2 (OpenAlex 2-year mean citedness, value as of 2026-10-31, retrieved 2026-11-01; not the official Clarivate IF)Reference: Basak S, Bera S, Misra SN, Rajan A, Poptani H, Chawla S, et al. Differentiating True Progression from Pseudoprogression in Glioblastomas: Dynamic Contrast Enhanced-MRI Habitat Analysis with Parsimonious Pharmacokinetic Model. Neuro-Oncology Advances. 2026 Oct 29 [Epub ahead of print]. doi:10.1093/noajnl/vdag254.Checked: Abstract only