NEUROBLASTOMA diagnosis and risk stratification could be improved through the use of a new multimodal artificial intelligence model capable of predicting both histological and molecular features from routine pathology data, according to findings from a large multi-institutional study involving 1,238 patients.
Addressing Challenges in Neuroblastoma Care
Neuroblastoma remains a leading cause of childhood cancer mortality and continues to present significant management challenges because of its biological heterogeneity and the limited availability of molecular profiling in routine clinical practice. To address these barriers, researchers developed NEVA, a multimodal foundation model designed specifically for neuroblastoma assessment.
Unlike conventional artificial intelligence approaches that rely on frozen encoders and multiple instance learning, the model was developed using a pathologist inspired hierarchical workflow combined with end-to-end optimisation. This design aimed to better replicate clinical decision-making processes while improving predictive performance across a range of clinically relevant tasks.
Neuroblastoma Prediction Across Multiple Clinical Tasks
The model was developed and evaluated using data from 1,238 patients collected across multiple institutions and centres. Performance was compared with 10 representative foundation models, including TITAN, UNI, and Virchow, across 11 separate clinical tasks.
NEVA outperformed comparator models across the majority of tasks evaluated. The model demonstrated strong diagnostic performance with data showing area under the receiver operating characteristic curve values of: 0.916 for subtype classification; 0.823 for Shimada classification; and 0.806 for risk group stratification.
In addition to diagnostic applications, the model predicted important molecular alterations directly from routinely available pathology material. Predictive performance reached an area under the receiver operating characteristic curve of: 0.924 for NMYC amplification; and 0.830 for 1p36 deletion.
Interpretable AI For Clinical Decision Support
The model also enabled prognostic stratification for progression free survival and overall survival across multiple test cohorts, highlighting its potential role in supporting treatment planning and patient management.
Importantly, NEVA incorporated interpretable attention maps capable of localising histologically relevant regions within pathology samples. Researchers suggested that this feature may improve transparency and confidence in model outputs while supporting clinical interpretation.
Taken together, the findings establish a scalable framework for neuroblastoma risk stratification and clinical decision support using routinely available pathology data.
Reference
Zhu J et al. A unified vision-language model for precision oncology and biomarker prediction in neuroblastoma. Nat Commun. 2026; https://doi.org/10.1038/ s41467-026-74865-5.
Featured image: Lexiconimages on Adobe Stock.
- Author:
