Machine learning algorithms and artificial neural networks for predicting schizophrenia using orbital parameters

Basit Öğe Kaydı
dc.contributor.authorEmre, Elif
dc.contributor.authorSoylemez, Derya Ozturk
dc.contributor.authorSecgin, Yusuf
dc.contributor.authorKaraagac, Seda Sogukpinar
dc.contributor.authorKenanoglu, Omer
dc.contributor.authorAydin, Suleyman
dc.date.accessioned2026-04-25T14:20:05Z
dc.date.available2026-04-25T14:20:05Z
dc.date.issued2025
dc.departmentSinop Üniversitesi
dc.description.abstractA persistent mental illness, schizophrenia has a complicated etiopathogenesis that includes both environmental and genetic elements. This study examined the possibility of diagnosing schizophrenia by utilizing computed tomography (CT) images of the orbit and its structures, which were then examined by artificial neural networks (ANNs) and machine learning (ML) algorithms. A retrospective analysis of the CT scans of 90 healthy people and 90 people with schizophrenia was conducted. Prior to measurement, all CT images underwent preprocessing steps to ensure align-ment and standardization. Height, width, depth, wall length, aperture area, interorbital width, biorbital width, bimalar width, skull transverse diameter, and optic nerve sheath width were among the orbital parameters that were measured. Statistical analysis revealed significant differences between the groups in left orbital width, left orbital aperture area, right optic nerve sheath width, transverse skull diameter, bimalar width, biorbital width, and left medial wall length. ML algorithms and ANNs were applied to the data, with the Extra Tree Classifier (ETC) algorithm achieving the highest accuracy of 0.78 and the Multilayer Perceptron Classifier (MLCP) model of ANN achieving an accuracy of 0.75 after 1000 training iterations. The Random Forest algorithm's SHAP analyzer determined that the left orbital width had the biggest impact on the final outcome. These results add to the expanding field of machine learning applications in psychiatry by indicating that AI-based models that analyze orbital morphometry may be useful instruments for detecting schizophrenia.
dc.identifier.doi10.1038/s41598-025-29610-1
dc.identifier.issn2045-2322
dc.identifier.issue1
dc.identifier.orcid0000-0003-1685-7802
dc.identifier.pmid41318706
dc.identifier.scopus2-s2.0-105026606747
dc.identifier.scopusqualityN/A
dc.identifier.urihttps://doi.org/10.1038/s41598-025-29610-1
dc.identifier.urihttps://hdl.handle.net/11486/8344
dc.identifier.volume16
dc.identifier.wosWOS:001654843700001
dc.identifier.wosqualityQ1
dc.indekslendigikaynakWeb of Science
dc.indekslendigikaynakScopus
dc.indekslendigikaynakPubMed
dc.language.isoen
dc.publisherNature Portfolio
dc.relation.ispartofScientific Reports
dc.relation.publicationcategoryMakale - Uluslararası Hakemli Dergi - Kurum Öğretim Elemanı
dc.rightsinfo:eu-repo/semantics/openAccess
dc.snmzKA_WOS_20260420
dc.subjectSchizophrenia
dc.subjectComputed tomography
dc.subjectMachine learning
dc.subjectArtificial neural networks
dc.subjectOrbital parameters
dc.titleMachine learning algorithms and artificial neural networks for predicting schizophrenia using orbital parameters
dc.typeArticle

Dosyalar

Koleksiyon

WoS İndeksli Yayınlar Koleksiyonu
PubMed İndeksli Yayınlar Koleksiyonu
Scopus İndeksli Yayınlar Koleksiyonu