Radiomics converts medical images into high-dimensional quantitative features that may capture subtle disease-related patterns not detectable through conventional analyses. The investigators pooled diagnostic performance with a random-effects meta-analysis and estimated summary sensitivity and specificity using a hierarchical summary receiver operating characteristic (HSROC) model. The researchers searched PubMed, EMBASE, IEEE Xplore and Scopus through September 17, 2026.

They followed PRISMA-DTA guidelines and used a preregistered protocol, identified as PROSPERO CRD420261286782. Risk of bias and reporting quality were assessed with QUADAS-2 and METRICS. Fourteen studies met the inclusion criteria, and eight contributed data to the main diagnostic meta-analysis.

For diagnostic discrimination, the pooled AUC was 0.81, with a 95% confidence interval of 0.73 to 0.88. The HSROC model estimated mean sensitivity at 0.736, with a 95% confidence interval of 0.666 to 0.795, and mean specificity at 0.735, with a 95% confidence interval of 0.668 to 0.793. Despite the strong overall AUC, statistical heterogeneity was substantial: the I-squared statistic was 84.2%.

This indicates that results varied markedly among the diagnostic studies contributing to the pooled estimate. The review characterized the studies’ overall quality as moderate to good, but its analysis also found that less rigorous validation procedures were associated with more optimistic performance. External validation was limited, and the reported aggregate results therefore do not establish that every radiomics model will perform equally well in routine clinical practice.

For treatment response, radiomics-based approaches also showed good discrimination in the meta-analysis. The reported AUC was 0.80 for longitudinal prediction and 0.88 for cross-sectional discrimination between treatment-response groups. These estimates concern the accuracy of statistical models in the included studies, not proof that radiomics can presently determine which treatment an individual patient should receive.

The supplied review abstract does not report confidence intervals or the number of contributing studies for the treatment-response estimates, limiting interpretation of their precision and evidence base. The authors called for standardized radiomics workflows, prospective study designs, integration with clinical markers and rigorous validation. Their findings position radiomics as a promising component of schizophrenia research for both diagnosis and treatment-response prediction, while showing that dependable individual-level clinical translation still requires stronger validation.

Subject of Research: Meta-analysis of radiomics-based MRI and PET models for schizophrenia diagnosis and treatment-response prediction

Article Title: Precision psychiatry in schizophrenia: a meta-analysis of radiomics-based advanced neuroimaging for diagnostic and treatment-response prediction

Article References: Caiazza, C., Fusco, G., Ugga, L., Toni, C., Cirillo, M., Sampogna, G., & Fiorillo, A. (2026). Precision psychiatry in schizophrenia: a meta-analysis of radiomics-based advanced neuroimaging for diagnostic and treatment-response prediction. Schizophrenia. https://doi.org/10.1038/s41537-026-00815-1

Image Credits: AI Generated

DOI: 10.1038/s41537-026-00815-1

Keywords: schizophrenia, radiomics, precision psychiatry, neuroimaging, MRI, PET, machine learning, treatment response, diagnostic accuracy, meta-analysis, Precision, psychiatry