Cancer imaging meeting maps MRI biomarkers from gliomas to myeloma

At the 25th International Cancer Imaging Society Meeting and Annual Teaching Course, held in Belfast, Northern Ireland, from 23 to 25 September 2026, researchers presented a broad sweep of imaging studies that push cancer diagnosis beyond simple tumour measurement. The meeting proceedings, published as open-access meeting abstracts in the journal Cancer Imaging, cover multiparametric MRI biomarkers in treated gliomas, MRI risk scoring for nerve tumours in children with neurofibromatosis type 1, whole-body MRI trajectories in multiple myeloma, and quantitative comparisons between emerging MRI techniques and conventional CT. Together, the abstracts describe a field moving from enhancement-based eyeballing toward structured, quantitative, and increasingly decision-oriented imaging.

One of the most detailed presentations came from Tata Medical Center in Kolkata, where Akshat Shah, Saugata Sen, and Aditi Chandra reported a retrospective study of 132 patients with treated diffuse gliomas who underwent serial post-treatment MRI between 2018 and 2025, generating 386 follow-up examinations. Each examination was assessed across six domains: enhancement evolution, T2/FLAIR change, diffusion abnormality, perfusion activity, MR spectroscopic abnormality, and temporal behaviour. The team classified the scans into five imaging-defined states: controlled treatment response, reactive treatment change, infiltrative progression, angiogenic progression, and necrotic-destabilised evolution. A composite multiparametric MRI index, with each domain scored from 0 to 3, was derived from these domains, and the reference outcome was established using longitudinal imaging, multidisciplinary assessment, treatment escalation, survival outcome, and histopathology where available.

The results suggest that this state-based framework captures information that conventional enhancement-based assessment misses. Controlled and reactive states were predominantly associated with stable disease or treatment-related change, whereas angiogenic and necrotic-destabilised states showed higher rates of confirmed progression and treatment escalation. The infiltrative state identified non-enhancing tumour evolution not adequately represented by enhancement alone. A multiparametric MRI index threshold of 10 or above identified probable progression with an area under the curve of 0.84, sensitivity of 81.6 percent, and specificity of 76.2 percent. Interobserver agreement for assigning imaging states was substantial, with a kappa of 0.72. The authors conclude that longitudinal multiparametric MRI biomarkers can map clinically relevant disease states in treated gliomas beyond conventional enhancement-based assessment and may improve the distinction between treatment effect and progression.

From Singapore, Timothy Shao Ern Tan and Jing Yuan of KK Women’s and Children’s Hospital presented the first validation of PLEXISS, a structured 10-point MRI risk score for plexiform neurofibromas in paediatric neurofibromatosis type 1. Malignant peripheral nerve sheath tumour is a major cause of disease-related mortality in these children, yet no validated MRI-based triage tool previously existed. The retrospective lesion-level study included 51 plexiform neurofibroma lesions from 31 paediatric patients at a tertiary children’s hospital. Scoring incorporated morphology from 0 to 4 points, diffusion-weighted imaging features from 0 to 4 points, and growth rate from 0 to 2 points, with predefined diffusion thresholds of ADC minimum at or below 1.1 times 10 to the minus 3 square millimetres per second and ADC mean at or below 1.8 times 10 to the minus 3. Malignancy was confirmed by histopathology or FDG-PET/CT where available, and remaining lesions were deemed non-malignant based on stable clinical and imaging follow-up.

Forty-four lesions, or 86 percent, were classified as low risk, five as intermediate risk, and two as high or very-high risk. Three lesions, or 6 percent, were malignant. Diffusion features and growth rate showed strong associations with risk category, with Spearman correlations of 0.707 and 0.545 respectively, both with p values below 0.001, whereas morphology alone was not significant. Malignant lesions showed significantly higher diffusion and growth scores, and diffusion-weighted imaging demonstrated the strongest discriminatory performance for malignancy, with an area under the curve of 0.809. The authors describe PLEXISS as showing promising preliminary validity as a practical MRI-based triage tool, with diffusion and interval growth emerging as the strongest predictors of malignant risk.

Several abstracts addressed myeloma imaging. A whole-body MRI study from the same Kolkata group followed 154 patients with confirmed multiple myeloma who underwent at least two whole-body MRI examinations between 2012 and 2024, yielding 426 examinations using a standardised protocol of T1-weighted Dixon, STIR, and diffusion-weighted imaging with ADC maps. Follow-up examinations were assessed according to MY-RADS response categories and classified into four longitudinal trajectories: improvement, stable low-risk, stable high-risk, and worsening. Improving trajectories had a median progression-free survival of 40 months, with overall survival not reached at five years, while stable low-risk trajectories showed less than 10 percent relapse at two years. Worsening trajectories showed median progression-free and overall survival of 11 and 20 months, and stable high-risk trajectories had the poorest outcome at 8 and 15 months. Survival differed significantly across groups, and trajectory remained independently associated with progression-free survival after adjustment for ISS stage and measurable residual disease status, with a hazard ratio of 0.42 and p below 0.001.

Also in myeloma, Fabian Bauer and colleagues at University Hospital Cologne evaluated a dual-energy CT-derived virtual non-calcium bone marrow biomarker for distinguishing monoclonal gammopathy of undetermined significance from newly diagnosed multiple myeloma. In 102 patients imaged between 2018 and 2025, the biomarker was significantly higher in myeloma than in MGUS, with mean values of 2.93 plus or minus 1.77 versus 1.38 plus or minus 0.95, and showed moderate correlation with bone marrow plasma cell infiltration across the entire cohort, with a Spearman coefficient of 0.38 and p below 0.001. The area under the curve for discriminating MGUS from myeloma was 0.78, with an optimal cutoff of 1.75. A companion study from the same group used automated spleen volumetry from routine baseline whole-body low-dose CT in 132 newly diagnosed myeloma patients and found that spleen volume correlated significantly with haemoglobin, M-protein, and revised International Staging System stage, with patients showing splenomegaly having lower haemoglobin and higher rates of stage 3 disease.

Structured reporting systems featured prominently. A retrospective quality-improvement study of 977 neuro-oncology MRI examinations discussed at multidisciplinary tumour board meetings between May 2024 and May 2026 found that implementing structured reporting with mandatory advanced MRI descriptors and standardised post-treatment terminology increased multidisciplinary treatment-decision concordance from 71 percent to 89 percent, while indeterminate recommendations requiring reassessment fell from 28 percent to 12 percent. Interobserver agreement improved from a kappa of 0.71 to 0.84. In a related comparison, BT-RADS outperformed RANO in 536 treated glioma patients, achieving 88 percent diagnostic accuracy for progression versus 77 percent, superior interobserver agreement, and clearer survival separation, with median progression-free and overall survival of 12 and 22 months for BT-RADS category 3 compared with 6 and 11 months for category 4.

Paediatric and gynaecological imaging also featured. A preliminary prospective study of 10 paediatric neuroblastoma patients found that non-contrast whole-body MRI achieved high concordance with MIBG scintigraphy, with a kappa of 0.82, sensitivity and specificity of 100 percent for primary tumour detection, and excellent agreement on tumour size, suggesting that gadolinium-based contrast may be omitted without loss of diagnostic performance. In ovarian imaging, a district general hospital review of 140 patients scored with O-RADS found that all histologically confirmed malignant lesions were classified as O-RADS 4 or 5 on both ultrasound and MRI, supporting accurate classification of high-risk pathology, though benign histology showed greater variability. A separate study of 159 patients with borderline ovarian tumours found that only 10 lesions, or 6 percent, recurred during five years of follow-up, and no MRI feature was predictive of recurrence.

The proceedings also include safety and technology evaluations. A meta-analysis of five cohort studies comprising 1,733 outpatients who underwent transthoracic lung biopsy found a pooled major complication rate of 7.6 percent, predominantly pneumothoraxes requiring chest drain insertion, a pooled minor complication rate of 28.1 percent, and a pooled readmission rate of 1.4 percent, with no procedure-related mortality. A prospective study of 40 patients compared zero-echo-time MRI with CT for metastatic bone lesions and found histogram area overlap of 93 percent for sclerotic lesions, 78 percent for lytic lesions, and 95 percent for mixed lesions, supporting zero-echo-time MRI as a clinically viable alternative to CT. In soft tissue sarcoma, a study of 36 patients with extremity undifferentiated pleomorphic sarcoma found that a model combining capsular enhancement, TIC-II, and favourable CE-SWI achieved an area under the curve of 0.98 in predicting pathology-assessed treatment effect, substantially outperforming RECIST, which achieved 0.52. Taken together, the Belfast proceedings illustrate a discipline in which imaging is increasingly asked not merely to detect tumours but to quantify biology, predict outcome, and directly inform treatment decisions across oncology.

Subject of Research: Multiparametric MRI and quantitative imaging biomarkers for cancer diagnosis, response assessment, and prognosis presented at the 25th International Cancer Imaging Society Meeting

Article Title: Proceedings of the 25th International Cancer Imaging Society Meeting and Annual Teaching Course

Article References: Proceedings of the 25th International Cancer Imaging Society Meeting and Annual Teaching Course. (2026). Cancer Imaging, 26(1), 125. https://doi.org/10.1186/s40644-026-01094-w (Author names are not included in the publisher’s citation export.)

Image Credits: AI Generated

DOI: 10.1186/s40644-026-01094-w

Keywords: cancer imaging, multiparametric MRI, glioma, multiple myeloma, neurofibromatosis type 1, radiomics, whole-body MRI, structured reporting, treatment response, prognosis, dual-energy CT, neuroblastoma