Abstract Introduction The benefits of artificial intelligence (AI) human partnerships—evaluating how AI agents enhance expert human performance—are increasingly studied. Though rarely evaluated in healthcare, an inverse approach is possible: AI benefiting from the support of an expert human agent. Brain tumour MRI characterisation, a technically challenging and specialist process which often necessitates gadolinium contrast administration, offers an ideal setting to evaluate such partnerships. Methods Here, we investigate both human-AI clinical partnership paradigms in the MRI-guided characterisation of patients with brain tumours. We utilised data from 11,089 patients with brain tumours across 10 datasets and 4 countries, evaluating 1,109 held-out test cases spanning 5 neuro-oncology categories: glioma, meningioma, metastases, paediatric lesions, and post-surgical appearances. Eleven consultant neuroradiologists across six London hospitals participated in a multi-reader, randomised crossover study. The task was to predict post-contrast tumour from pre-contrast MRI sequences alone, with and without the corresponding agent. Results We reveal that human-AI partnerships improve accuracy and metacognitive ability not only for radiologists supported by AI, but also for AI agents supported by radiologists. AI supported by radiologists achieved the highest balanced accuracy (0.840), followed by AI alone (0.824), radiologists with AI support (0.743), and radiologists alone (0.698). Inter-rater agreement significantly improved with agent support (Cohen’s kappa 0.338 to 0.484; p < 0.0001). The greatest patient benefit was evident with an AI agent supported by a human one. Synergistic improvements in agent accuracy, metacognitive performance, and inter-rater agreement suggest that AI can create more capable, confident, and consistent clinical agents, whether human or model-based. Conclusions Our work suggests that the maximal value of AI in neuro-oncology and beyond could emerge not from replacing human intelligence, but from AI agents that leverage and amplify it.
Abstract Introduction Teenage and Young Adult (TYA) patients aged 16–24 years represent a biologically and clinically distinct cohort of central nervous system (CNS) tumour patients. Despite national standards for TYA cancer care, no dedicated management pathway for TYA brain tumours exists at Trust or national level. We developed a Trust-wide Standard Operating Procedure (SOP) for TYA CNS tumours at University Hospitals Birmingham (UHB), alongside an artificial intelligence (AI)-assisted framework for structured SOP development and optimisation. Methods Literature review identified no Trust-specific or UK national guidance dedicated to TYA brain tumour pathways. A draft SOP was developed using ChatGPT-5, a large language model (LLM), guided by structured prompts referencing SIOP Europe Brain Tumour Group consensus guidance, NICE standards, and NHS England TYA service specifications. The LLM supported drafting and formatting but did not replace clinical oversight. Multidisciplinary consensus was achieved using a modified Digital Delphi methodology with anonymous 5-point Likert voting. The predefined scope included referral, investigation, MDT processes, surgical and oncological pathways, tumour-specific follow-up, and late-effects surveillance for patients aged 16–24 years. Consensus was defined as ≥ 80% agreement. In parallel, we developed an SOP Quality Assessment Tool and applied readability analysis (Flesch-Kincaid, SMOG, Gunning Fog) to existing Neurosurgery SOPs. LLM-assisted prompt engineering was used to reformat narrative-heavy SOPs into structured, checklist-driven templates, which were then applied to the TYA pathway. Results A consensus-based Trust-wide SOP was developed and ratified. AI-assisted reformatting improved structural clarity and usability, enabling development of a standardised template for TYA brain tumour management. Conclusions There is a clear gap in guidance for TYA brain tumour pathways. We demonstrate a scalable, consensus-driven, AI-assisted model for SOP development that could support standardisation of TYA neurooncology care regionally and nationally.
Abstract Introduction Rapid diagnostic tools could improve the early detection of glioblastoma (GBM) and brain metastases (BM). Lateral flow assays (LFAs) offer a simple, point-of-care testing platform. Glial fibrillary acidic protein (GFAP) has previously been reported as a circulating marker associated with blood–brain barrier disruption in brain tumours. This study aimed to develop and evaluate an LFA capable of detecting GFAP in the serum of patients with GBM and BM. Methods Detection antibodies were conjugated to universal gold nanoparticles to generate a sandwich-format GFAP LFA test strip. The assay was evaluated using serum samples from patients with GBM (n = 20), BM (n = 12), and non-tumour controls (n = 30). Test line signal intensity was quantified using Fiji/ImageJ software. Serum GFAP concentrations were also measured using Single Molecule Array (SIMOA®) technology as a reference method. Results Positive LFA test bands were observed in 95% of GBM samples (19/20), 83% of BM samples (10/12), and 30% of control samples (10/30). Mean test band intensities were 11,408 arbitrary units (a.u.) for GBM, 11,206 a.u. for BM, and 14,072 a.u. for controls, indicating an approximate false-positive rate of 30%. SIMOA® analysis detected GFAP in all samples across all groups, with mean serum concentrations of 2,533 pg/mL in GBM, 644 pg/mL in BM, and 45 pg/mL in controls. Conclusion This study demonstrates the successful development of a highly sensitive GFAP lateral flow assay for detecting blood–brain barrier disruption in GBM and BM. While the assay shows strong sensitivity, a false-positive rate remains and will be addressed through further optimisation in future work.
Abstract Introduction Integrated histopathological-molecular analysis is essential to diagnose and guide precision therapies in brain tumours. Whilst standard of care (SoC) diagnostics takes weeks, long read sequencing (LRS) on Oxford Nanopore Technology (ONT) platform and bioinformatics research tools provide rapid diagnosis within hours/days. We independently replicated ONT-LRS diagnostic assay within the NHS. Methods Patients undergoing brain tumour surgery were prospectively consented for surplus tissue sampling for research. LRS was performed on 54 fresh tissue and 78 fresh frozen tissue specimens. University of Birmingham scientists trained NHS staff to extract tumour DNA for sequencing using ONT PromethION flow cells. Research tools ROBIN, EPINN and MethyLYZR were adopted for real-time and batch methylome array. Results 132 specimens from 112 patients were analysed: mean age was 53(16.8); 70(63%) were male; self-reported ethnicity was White(76.5%), Asian(16.2%), Black(2.7%) or other(4.5%). Integrated SoC confirmed Glioblastoma IDH-wild type(59.5%), Astrocytoma IDH-mutant(27%), Oligodendroglioma IDH-mutant, 1p19q-codeleted(12.6%). Diffuse hemispheric glioma(H3G34-mutant; 0.9%), Anaplastic Pleomorphic Xanthoastrocytoma(0.9%), Subependymal gliant cell astrocytoma(0.9%), Pilocytic astrocytoma(0.9%), Meningioma(0.9%), Metastasis (Melanoma, Oesophageal, Renal, 2.7%). Integrating LRS within the SoC neuropathology workflow demonstrated 100% feasibility with no failures in tissue handling, DNA extraction, or sequencing. Amongst prospectively recruited patients(n = 54), NHS SoC integrated diagnosis was reported a median (IQR) of 23(18,32) days after surgery, whilst ROBIN obtained a methylome-based diagnosis within 1.4 days from surgery. In the total population (n = 132), LRS methylome-based diagnostic concordance was 77.6% for EPINN, 75.2% ROBIN and 56% MethyLYZR. Discordant samples included: tumours outside classifier remit(n = 3) and low tumour purity cases(n = 5). Conclusion This demonstrates good concordance between SoC and DNA LRS results. Building on close working partnership between surgeons, pathologists and scientists, we need access to larger, diverse sample populations to refine classifiers and understand their limitations.
Abstract Introduction The National Genomic Test Directory (NGTD), introduced in 2019, defines NHS commissioned genomic tests in England and outlines their clinical eligibility. In cancers of the central nervous system (CNS), neuropathologists routinely integrate genomic analysis into diagnostic workflows to support tumour classification and clinical patient management. However, the existing Excel based NGTD is difficult to navigate and lacks the flexibility required, highlighting the need for a modernised, digital alternative. Methods The NHS England (NHSE) Genomics Unit undertook extensive stakeholder engagement to define the requirements for a digital directory. Test Directory content was evaluated alongside Patient Level Contract Monitoring data to identify requesting patterns, redundancy and evolving clinical needs. The WHO Blue Book on CNS Tumours was reviewed with adult and paediatric neuropathology experts to determine diagnostic genomic testing requirements, and relevant precision medicine genomic targets. The Digital Genomic Test Directory Service (DGTS) was developed as the platform to deliver the NGTD and support transformation of genomic testing for CNS tumours. Results Stakeholder input and data review demonstrated the need to update and restructure the NGTD in alignment with WHO classifications. A differential diagnostic approach was adopted, incorporating broader NGS panels that support integrated diagnosis while retaining targeted assays for urgent or salvage testing. Redundant tests identified through analysis were removed. Phase 1 of the programme has been completed, consolidating CNS tumour indications from 58 indications into seven streamlined test packages. Conclusions NHSE has initiated a major transformation to deliver the NGTD in a fully digital format via the DGTS, improving access to commissioned genomic tests essential for CNS tumour diagnosis and management. CNS tumours represent the first transformed disease area, setting the foundation for broader digital transformation across solid cancers.
