Systematic review of AI for schizophrenia prediction using medical imaging data

Photo by Zach M / Unsplash

Frontiers in Medicine Published May 17, 2026 Medically reviewed May 17, 2026 Study authors: Suresh Babu Palpandi, Nagaraj Palanigurupackiam, Hessa Almatar, Reema Alduhayan, Barrak Alsomaie, Ah… DOI ↗ By Dr. Sofia Müller, MD · Lifespan & Whole-Person Care

Key Takeaway

Note that AI models for schizophrenia prediction face significant challenges and limitations.

This systematic literature review evaluates 185 studies that utilize artificial intelligence methodologies, specifically machine learning and deep learning, in conjunction with medical imaging data including MRI, PET, and EEG. The scope of the review encompasses the prediction of schizophrenia, biomarker identification, model selection, and diagnostic accuracy within this technological context. The authors do not report specific adverse events or tolerability data as these were not included in the source material.

The synthesis of these 185 studies reveals significant challenges and limitations inherent to machine learning and deep learning models when applied to psychiatric imaging. The review does not establish causal relationships or provide specific numerical effect sizes, as the primary results were not reported in the source data. Instead, the authors focus on the methodological hurdles and the current state of research gaps in this field.

The practice relevance identified by the authors suggests that these technologies may support timely interventions for individuals affected by schizophrenia. However, the review does not provide definitive clinical guidelines or specific safety profiles. The certainty of these conclusions is constrained by the nature of the literature review and the lack of reported follow-up durations or specific study settings.

Study Details

Study typeSystematic review

EvidenceLevel 1

PublishedMay 2026

View Original Abstract ↓

Schizophrenia (SZ) is a debilitating mental illness that adversely affects social and family interactions, ranking as a leading contributor to global disability. Existing diagnostic approaches, including MRI, PET, and EEG, underscore the necessity for effective predictive strategies to enhance management and reduce costs. This review evaluates the application of artificial intelligence (AI) methodologies—specifically Machine Learning (ML) and Deep Learning (DL) in predicting SZ using medical imaging data, while addressing existing challenges and identifying key biomarkers to improve diagnostic accuracy. A systematic literature review was performed using the databases IEEE, PubMed, ScienceDirect, MDPI, Google Scholar, and Springer from inception until March 31, 2026. The initial search generated 820 records, and after a thorough screening process, 185 studies relevant to disease diagnosis, model selection across various neuroimaging modalities, including biomarker identification, were identified. The review protocol has been registered with PROSPORO registration: CRD420251131635. The studies were selected based on different medical imaging data related to SZ. This review presents a thorough examination of advancements in SZ detection via AI methodologies. It highlights not only providing existing predictive techniques, identifies research gaps, biomarkers identification and assessment, and underscores the potential of AI-based ML and DL methods to facilitate early and accurate diagnosis of SZ. Five unimodal and various combinations of multimodal data were examined, along with the AI models’ performance metrics from multiple studies. The review provides a comprehensive assessment of AI algorithms relevant to both unimodal and multimodal data, biomarkers of neuroimaging modalities with ROIs, challenges, and limitations of ML and DL models, and future directions of prediction for clinical diagnosis, thereby supporting timely interventions for individuals affected by SZ. https://www.crd.york.ac.uk/PROSPERO/view, identifier CRD420251131635.