Alessandro Pigoni, Isidora Tesic, Cecilia Pini, Paolo Enrico, Lorena Di Consoli, Francesca Siri, Guido Nosari, Adele Ferro, Letizia Squarcina, Giuseppe Delvecchio, Paolo Brambilla
{"title":"Multimodal Machine Learning Prediction of 12-Month Suicide Attempts in Bipolar Disorder.","authors":"Alessandro Pigoni, Isidora Tesic, Cecilia Pini, Paolo Enrico, Lorena Di Consoli, Francesca Siri, Guido Nosari, Adele Ferro, Letizia Squarcina, Giuseppe Delvecchio, Paolo Brambilla","doi":"10.1111/bdi.70011","DOIUrl":null,"url":null,"abstract":"<p><strong>Introduction: </strong>Bipolar disorder (BD) patients present an increased risk of suicide attempts. Most current machine learning (ML) studies predicting suicide attempts are cross-sectional, do not employ time-dependent variables, and do not assess more than one modality. Therefore, we aimed to predict 12-month suicide attempts in a sample of BD patients, using clinical and brain imaging data.</p><p><strong>Methods: </strong>A sample of 163 BD patients were recruited and followed up for 12 months. Gray matter volumes and cortical thickness were extracted from the T1-weighted images. Based on previous literature, we extracted 56 clinical and demographic features from digital health records. Support Vector Machine was used to differentiate BD subjects who attempted suicide. First, we explored single modality prediction (clinical features, GM, and thickness). Second, we implemented a multimodal stacking-based data fusion framework.</p><p><strong>Results: </strong>During the 12 months, 6.13% of patients attempted suicide. The unimodal classifier based on clinical data reached an area under the curve (AUC) of 0.83 and balanced accuracy (BAC) of 72.7%. The model based on GM reached an AUC of 0.86 and BAC of 76.4%. The multimodal classifier (clinical + GM) reached an AUC of 0.88 and BAC of 83.4%, significantly increasing the sensitivity. The most important features were related to suicide attempts history, medications, comorbidities, and depressive polarity. In the GM model, the most relevant features mapped in the frontal, temporal, and cerebellar regions.</p><p><strong>Conclusions: </strong>By combining models, we increased the detection of suicide attempts, reaching a sensitivity of 80%. Combining more than one modality proved a valid method to overcome limitations from single-modality models and increasing overall accuracy.</p>","PeriodicalId":8959,"journal":{"name":"Bipolar Disorders","volume":" ","pages":""},"PeriodicalIF":5.0000,"publicationDate":"2025-03-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Bipolar Disorders","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1111/bdi.70011","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CLINICAL NEUROLOGY","Score":null,"Total":0}
引用次数: 0
Abstract
Introduction: Bipolar disorder (BD) patients present an increased risk of suicide attempts. Most current machine learning (ML) studies predicting suicide attempts are cross-sectional, do not employ time-dependent variables, and do not assess more than one modality. Therefore, we aimed to predict 12-month suicide attempts in a sample of BD patients, using clinical and brain imaging data.
Methods: A sample of 163 BD patients were recruited and followed up for 12 months. Gray matter volumes and cortical thickness were extracted from the T1-weighted images. Based on previous literature, we extracted 56 clinical and demographic features from digital health records. Support Vector Machine was used to differentiate BD subjects who attempted suicide. First, we explored single modality prediction (clinical features, GM, and thickness). Second, we implemented a multimodal stacking-based data fusion framework.
Results: During the 12 months, 6.13% of patients attempted suicide. The unimodal classifier based on clinical data reached an area under the curve (AUC) of 0.83 and balanced accuracy (BAC) of 72.7%. The model based on GM reached an AUC of 0.86 and BAC of 76.4%. The multimodal classifier (clinical + GM) reached an AUC of 0.88 and BAC of 83.4%, significantly increasing the sensitivity. The most important features were related to suicide attempts history, medications, comorbidities, and depressive polarity. In the GM model, the most relevant features mapped in the frontal, temporal, and cerebellar regions.
Conclusions: By combining models, we increased the detection of suicide attempts, reaching a sensitivity of 80%. Combining more than one modality proved a valid method to overcome limitations from single-modality models and increasing overall accuracy.
期刊介绍:
Bipolar Disorders is an international journal that publishes all research of relevance for the basic mechanisms, clinical aspects, or treatment of bipolar disorders and related illnesses. It intends to provide a single international outlet for new research in this area and covers research in the following areas:
biochemistry
physiology
neuropsychopharmacology
neuroanatomy
neuropathology
genetics
brain imaging
epidemiology
phenomenology
clinical aspects
and therapeutics of bipolar disorders
Bipolar Disorders also contains papers that form the development of new therapeutic strategies for these disorders as well as papers on the topics of schizoaffective disorders, and depressive disorders as these can be cyclic disorders with areas of overlap with bipolar disorders.
The journal will consider for publication submissions within the domain of: Perspectives, Research Articles, Correspondence, Clinical Corner, and Reflections. Within these there are a number of types of articles: invited editorials, debates, review articles, original articles, commentaries, letters to the editors, clinical conundrums, clinical curiosities, clinical care, and musings.