Advancement of deep learning models with whole slide image in diagnosis, subtyping and prognosis for glioma.

Clinical Snapshot

Key Findings

• P Value: Not reported in the abstract or available metadata

• Effect Size: Reported AUC values: ResNet-50 (glioma-specific) AUC 0.983 for subtype classification; Vision Transformer AUC 0.960 for molecular typing; ROAM (hybrid) AUC 0.990; CHIEF (pan-cancer hybrid) AUC 0.9397 for diagnosis

• Primary Outcome: Diagnostic, subtyping, grading, and prognostic performance of DL models applied to glioma WSIs, reported as AUC across multiple architectures and datasets

• Nnt Or Sensitivity: Sensitivity and specificity not reported in the abstract; no NNT calculable; AUC values serve as the primary performance metric but are not accompanied by operating point-specific sensitivity/specificity data

• Confidence Interval: Not reported for any cited metric

Clinical Application

Clinical implementation requires significant infrastructure investment: whole slide scanners, high-performance computing, validated DL software, and integration with laboratory information systems. Regulatory approval (TGA in Australia) for diagnostic DL tools in neuropathology is not yet established for most described models. Workflow integration with existing neuropathology reporting remains a substantial barrier. Lightweight model development, noted as a future priority, may eventually facilitate broader deployment. Australia's neuropathology workforce is concentrated in major metropolitan centres, and digital pathology adoption is increasing but uneven across states. No TGA-approved DL diagnostic tool for glioma WSI analysis is currently listed. The RACGP and RANZCR have not yet issued specific guidelines on DL-assisted neuropathology. PBS implications are indirect — improved molecular subtyping could influence eligibility for targeted therapies (e.g., IDH-mutant glioma management). Australian glioma registries (e.g., through ANZSNSG) could provide valuable external validation datasets. The predominantly TCGA/CPTAC-derived evidence base may not fully reflect Australian patient demographics or tissue processing standards used in NATA-accredited laboratories. Adult patients with suspected or confirmed glioma undergoing surgical resection or biopsy with available digitised histopathological specimens; most relevant to tertiary neuro-oncology centres with digital pathology infrastructure

Abstract

Gliomas are heterogeneous primary central nervous system tumors with diverse molecular subtypes and variable prognosis. A paradigm shift in histopathology is underway as samples are digitized into whole-slide images (WSIs). Deep learning (DL) shows great potential for glioma diagnosis, subtyping, grading, and prognosis. This review summarizes recent advances in WSI-based DL models, covering data preprocessing, model architectures, and translational performance. Model evolution has progressed from CNNs (e.g. ResNet, capturing local features via convolution) to Transformers (e.g. Vision transformer, modeling global dependencies via self-attention), hybrid architectures, and large language model. Typical pipelines include quality control, stain normalization, patch extraction, feature integration, and prediction. Public datasets such as The Cancer Genome Atlas and Clinical Proteomic Tumor Analysis Consortium serve as key resources. On various datasets (public and institutional), a glioma-specific model ResNet-50 achieved AUC 0.983 for subtype classification; a Vision Transformer model reached AUC 0.960 for molecular typing; a hybrid model, ROAM, attained AUC 0.990; and the pan-cancer hybrid model called CHIEF achieved AUC 0.9397 for diagnosis. Attention heatmaps and Shapley Additive exPlanations provide interpretability by linking model outputs to histologic regions. Novel multimodal fusion integrates genomic, proteomic, radiologic, and clinical data to enhance prediction and uncover biological relevance. Future research may focus on enhancing model generalizability and predictive accuracy via advanced architectures, developing lightweight models for resource-limited settings to facilitate clinical translation.

References

1. 1.Gui, K., Li, Q., Xie, Z., Zhan, T., Long, X., Yang, G., Luo, K., & Liu, H. (2026). Advancement of deep learning models with whole slide image in diagnosis, subtyping and prognosis for glioma. Progress in Biomedical Engineering. https://doi.org/10.1088/2516-1091/ae7860