Artificial Intelligence for Diagnosis, Risk Stratification, and Prognosis of Neuroblastoma - A Systematic Review and Meta-Analysis.

What was studied

This was a review and meta-analysis of artificial intelligence (AI) in neuroblastoma (NB). It looked at diagnosis, risk groups, outlook and gene features. The authors searched several databases. The authors pooled results with random-effects models. Of 3,742 articles found, 53 were included. The models were machine learning models or hybrid nomograms.

What they found

Machine learning models scored higher than radiologists at telling diagnoses apart. The AUC was 0.87 against 0.83. The gap was not significant. Hybrid nomograms reached an AUC of 0.87 for risk groups. AI nomograms reached an AUC of 0.9 and gene signatures 0.8. Both beat usual markers, in a descriptive way. Predicting response to chemo stayed low. Only some models reported calibration or outside validation.

What it means for patients

The authors say AI shows proof of concept across NB care. They say use in clinics is too early. They say new studies should test the models ahead of time in child cohorts.

Why it is not certain yet

The gap with radiologists was very uncertain. Only some models reported calibration or outside validation. The authors say dataset size and child-specific model work fell short. They say studies should use shared definitions.

Source

Artificial Intelligence for Diagnosis, Risk Stratification, and Prognosis of Neuroblastoma - A Systematic Review and Meta-Analysis.

We restate facts from the source record in our own words. The original wording belongs to its publisher.