基于影像组学及机器学习预测个体首发精神分裂症患者抗精神病药物治疗反应性的磁共振研究

Precision medicine, the image first. Whether the use of the characteristic biomarkers of schizophrenia could achieve early diagnosis and treatment of diseases is one of the hot issues in global medical research. Although previous imaging studies have made a significant contribution to the understanding of the etiology and pathogenesis of schizophrenia, the clinical application value is limited. The main reason is that previous studies usually compare differences between groups, rather than to explore the diagnosis and treatment of patients with schizophrenia at the individual level. This project is based on the first-episode never-treated schizophrenia patients, focusing on the key scientific question of how to individualize the prediction of the antipsychotic drug treatment response in patients with first-episode schizophrenia. Based on the series of research results the applicant's found when analyzing the brain mechanism of schizophrenia, combined with multi-dimensional symptom information and multi-modal magnetic resonance imaging technology, using radiomics and the latest methods of machine learning, the applicant will systematically explore the mechanism of different responsiveness of drug treatment in different patients and the feasibility of individualized predictive analysis. The successful development of this project and the establishment of a stable and effective predictive model will help to minimize patient suffering and maximize the effective allocation of health care resources, which with high value in translational medicine and health economics.

精准医疗，影像先行。能否利用精神分裂症特征性生物学标志物实现疾病的早期精准诊治，是当前全球医学研究的热点问题之一。尽管以往的影像学研究对精神分裂症的病因和发病机制的深入理解做出了巨大的贡献，但临床应用价值有限。原因在于以往研究通常比较组间差异，而非个体水平探索精神分裂症患者的诊治。本项目以首发未治疗精神分裂症患者为研究对象，围绕“如何个体化预测首发精神分裂症患者抗精神病药物治疗反应性”这一关键科学问题，基于申请人解析精神分裂症疾病相关脑机制的系列研究成果，结合多维症状信息和多模态磁共振成像技术，运用影像组学分析和机器学习最新方法，系统探讨不同患者药物治疗不同反应性的机制及个体化预测分析的可行性。本项目的成功开展和稳定有效的预测模型的建立，有助于最小化患者痛苦和最大化医疗卫生资源的有效分配，具有较高的转化医学和卫生经济学价值。

精神分裂症的病因复杂，病人的临床表现和预后异质性较高，亟需可靠的生物学标志物实现疾病的早期精准诊治。本项目基于大样本精神分裂症人群，结合多维症状信息和多模态磁共振成像技术，运用影像组学分析和机器学习最新方法，实现了精神分裂症患者的异质性解析和个体化分类预测。主要包括：1）基于影像特征驱动的聚类分析方法，从精神影像生物学的角度揭示了精神分裂症存在疾病异质性；2）解析出精神分裂症不同患者亚型，以及不同亚型患者相对应的脑解剖-行为或症状耦合方式，为精神分裂症复杂的临床表现提供新的解释和生物学证据；3）构建了基于脑灰质结构特征的影像组学模型，辅助精神分裂症患者进行个体化的分类诊断及抗精神病药物治疗反应性的预测。综上，本项目运用了机器学习和影像组学技术，探索了精神分裂症患者临床异质性的机制，构建了稳定有效的个体化诊断预测模型，有助于最小化患者痛苦和最大化医疗卫生资源的有效分配，具有较高的转化医学和卫生经济学价值。