Frontiers in psychiatry · 2025
Support vector machine-based classification of bulimia nervosa using diffusion tensor imaging.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Zheng Linli, Wang Yu, Jing Ma, Wang Meiou, Liu Yang, Li Jin, Li Tao, Zhang Lan
Аннотация
BACKGROUND: Alterations in brain structure have been suggested to be associated with bulimia nervosa (BN). This study aimed to employ machine learning (ML) methods based on diffusion tensor imaging (DTI) to facilitate the diagnosis of BN and to identify potential neurobiological markers.
METHODS: Thirty-four drug-naive females with bulimia nervosa (BN) and 34 age- and gender-matched healthy controls (HCs) underwent diffusion tensor imaging (DTI) scanning. The extracted features included fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD), and mean diffusivity (MD). Support vector machines (SVM), a commonly used machine learning (ML) approach, were employed to distinguish individuals with BN from healthy controls.
RESULTS: Five ML models were constructed. The FA model (AUC=0.821) and the combined FA+MD+AD+RD model (AUC=0.739) exhibited satisfactory classification performance, with the FA model exhibiting the most effective results. The FA model achieved an accuracy of 82.35%, a specificity of 82.35%, and a sensitivity of 85.29%. The contributing brain regions were primarily located in the frontal lobe, brainstem, temporal lobe, midbrain, cerebellar tonsil, and posterior cerebellar lobe. In contrast, the MD model (AUC=0.689), the AD model (AUC=0.621), and the RD model (AUC=0.625) demonstrated poor classification performance.
CONCLUSIONS: This study demonstrated that DTI-based machine learning (ML) approaches could effectively differentiate individuals with bulimia nervosa (BN) from healthy controls (HCs), thereby providing insights into potential neurobiological markers associated with BN.
METHODS: Thirty-four drug-naive females with bulimia nervosa (BN) and 34 age- and gender-matched healthy controls (HCs) underwent diffusion tensor imaging (DTI) scanning. The extracted features included fractional anisotropy (FA), axial diffusivity (AD), radial diffusivity (RD), and mean diffusivity (MD). Support vector machines (SVM), a commonly used machine learning (ML) approach, were employed to distinguish individuals with BN from healthy controls.
RESULTS: Five ML models were constructed. The FA model (AUC=0.821) and the combined FA+MD+AD+RD model (AUC=0.739) exhibited satisfactory classification performance, with the FA model exhibiting the most effective results. The FA model achieved an accuracy of 82.35%, a specificity of 82.35%, and a sensitivity of 85.29%. The contributing brain regions were primarily located in the frontal lobe, brainstem, temporal lobe, midbrain, cerebellar tonsil, and posterior cerebellar lobe. In contrast, the MD model (AUC=0.689), the AD model (AUC=0.621), and the RD model (AUC=0.625) demonstrated poor classification performance.
CONCLUSIONS: This study demonstrated that DTI-based machine learning (ML) approaches could effectively differentiate individuals with bulimia nervosa (BN) from healthy controls (HCs), thereby providing insights into potential neurobiological markers associated with BN.