Laboratory of Molecular Psychiatry, Hospital de Clínicas de Porto Alegre, Avenida Ramiro Barcelos, 2350, 90035-903, Porto Alegre, Brazil; Programa de Pós-Graduação em Psiquiatria e Ciências do Comportamento, Universidade Federal do Rio Grande do Sul, Rua Ramiro Barcelos, 2400 - 2° andar, 90035-003, Porto Alegre, Brazil.
Departamento de Psiquiatria, Universidade Federal do Paraná, Rua Padre Camargo, 280 - 6º andar, 80060-240, Curitiba, Brazil.
J Psychiatr Res. 2018 Aug;103:237-243. doi: 10.1016/j.jpsychires.2018.05.023. Epub 2018 May 26.
Neuroimaging studies have been steadily explored in Bipolar Disorder (BD) in the last decades. Neuroanatomical changes tend to be more pronounced in patients with repeated episodes. Although the role of such changes in cognition and memory is well established, daily-life functioning impairments bulge among the consequences of the proposed progression. The objective of this study was to analyze MRI volumetric modifications in BD and healthy controls (HC) as possible predictors of daily-life functioning through a machine learning approach. Ninety-four participants (35 DSM-IV BD type I and 59 HC) underwent clinical and functioning assessments, and structural MRI. Functioning was assessed using the Functioning Assessment Short Test (FAST). The machine learning analysis was used to identify possible candidates of regional brain volumes that could predict functioning status, through a support vector regression algorithm. Patients with BD and HC did not differ in age, education and marital status. There were significant differences between groups in gender, BMI, FAST score, and employment status. There was significant correlation between observed and predicted FAST score for patients with BD, but not for controls. According to the model, the brain structures volumes that could predict FAST scores were: left superior frontal cortex, left rostral medial frontal cortex, right white matter total volume and right lateral ventricle volume. The machine learning approach demonstrated that brain volume changes in MRI were predictors of FAST score in patients with BD and could identify specific brain areas related to functioning impairment.
在过去的几十年里，神经影像学研究一直在双相情感障碍（BD）中不断探索。神经解剖学的变化在反复发作的患者中更为明显。尽管这种变化在认知和记忆中的作用已得到充分证实，但在提出的进展中，日常生活功能障碍的影响更为突出。本研究的目的是通过机器学习方法分析 BD 和健康对照组（HC）的 MRI 容积变化，作为日常生活功能的可能预测指标。94 名参与者（35 名 DSM-IV BD 型 I 和 59 名 HC）接受了临床和功能评估以及结构 MRI。使用功能评估简短测试（FAST）评估功能。通过支持向量回归算法，机器学习分析用于识别可能的局部脑容量候选者，以预测功能状态。BD 患者和 HC 在年龄、教育程度和婚姻状况方面没有差异。两组在性别、BMI、FAST 评分和就业状况方面存在显著差异。BD 患者的观察到的和预测的 FAST 评分之间存在显著相关性，但对照组则没有。根据该模型，能够预测 FAST 评分的脑结构体积包括：左侧额上回、左侧额极内侧回、右侧白质总体积和右侧侧脑室体积。机器学习方法表明，MRI 中的脑体积变化是 BD 患者 FAST 评分的预测指标，并可以识别与功能障碍相关的特定脑区。
