Back to all papers

Artificial Intelligence and CT Neuroimaging in Dementia and Psychotic Disorders: A Viewpoint.

February 25, 2026pubmed logopapers

Authors

Bampton TJ,Weber A,Bastiampillai T,Palmer LJ,Clark S

Affiliations (10)

  • Discipline of Psychiatry, Adelaide Medical School, College of Health, Adelaide University, Adelaide, Australia. [email protected].
  • Central Adelaide Local Health Network, Adelaide, Australia. [email protected].
  • Australian Institute for Machine Learning, Adelaide University, Adelaide, Australia. [email protected].
  • Central Adelaide Local Health Network, Adelaide, Australia.
  • Adelaide Medical School, Adelaide University, Adelaide, Australia.
  • Southern Adelaide Local Health Network, Adelaide, Australia.
  • Flinders University, Adelaide, Australia.
  • Australian Institute for Machine Learning, Adelaide University, Adelaide, Australia.
  • School of Public Health, Adelaide University, Adelaide, Australia.
  • Discipline of Psychiatry, Adelaide Medical School, College of Health, Adelaide University, Adelaide, Australia.

Abstract

Psychotic disorders are marked by heterogeneity in symptoms and treatment response, yet efforts to develop clinically useful predictive models through neuroimaging have been limited. Although MRI is gold standard for neuroimaging in psychosis, CT is frequently used in psychiatric care due to accessibility and cost. Advances in artificial intelligence, particularly deep learning, offer the potential for extracting meaningful information from CT brain scans. We conducted a systematic review of peer-reviewed studies applying machine learning, including deep learning to CT brain imaging in dementia or psychotic disorders. Six databases were searched from inception to Nov. 22, 2024. Studies were included if they involved deep or machine learning models using CT scans for diagnostic or prognostic modelling. Risk of bias was assessed using PROBAST + AI; reporting quality was evaluated with TRIPOD-AI or CLAIM. Seven studies met inclusion criteria: one related to psychosis and six to dementia. All studies included used deep learning models for classification or segmentation. Sample sizes ranged from 65 to 917. All predictive studies were rated as high risk of bias. TRIPOD-AI adherence was limited with most items poorly reported (16-21/27). Segmentation studies had low-to-moderate CLAIM adherence with limitations in external validation, code availability and reproducibility. Only one study to date has applied deep learning to CT imaging in psychosis. Research in dementia offers transferable insights. From a translational viewpoint, future work should prioritise the construction of large and well characterised psychosis cohorts including medical imaging, clinical and functional outcome prediction, external validation and transparent methodology to support clinical translation.

Topics

Journal ArticleReview

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.