女性快感缺失型抑郁症患者的语音特征分析

Analysis of speech features in female depression patients with anhedonia symptoms

目的探讨女性快感缺失型抑郁症患者的语音特征,及其对快感缺失症状的识别作用。方法选取2020年9月至2021年10月在南京脑科医院住院的女性抑郁症患者102例,其中62例快感缺失型抑郁症患者(快感缺失组),40例非快感缺失型抑郁症患者(非快感缺失组)。同期招募女性健康对照50例。所有被试均进行17项汉密尔顿抑郁量表(17-item Hamilton depression scale,HAMD-17)、斯奈斯汉密尔顿愉快体验量表(Snaith-Hamilton pleasure scale,SHAPS)和时间性愉快体验量表(the temporal experience of pleasure scale,TEPS)评估,并进行语音采集。采用SPSS 23.0软件进行数据处理。统计方法采用单因素方差分析、非参数检验、Logistic回归与受试者工作特征曲线分析。结果与非快感缺失组相比,快感缺失组患者梅尔倒谱系数、共振峰、声强和能量特征共有15个语音特征存在显著改变(均P<0.05);其中梅尔倒谱系数特征识别女性快感缺失型抑郁症的准确率最高,敏感度47.5%,特异度91.9%,曲线下面积(area under curve,AUC)为0.751,95%CI=0.686~0.866;共振峰特征在识别女性快感缺失型抑郁症中,敏感度90.0%,特异度40.3%,AUC为0.647,95%CI=0.605~0.824;能量特征在识别女性快感缺失型抑郁症中,AUC为0.679,敏感度60.0%,特异度74.2%,95%CI=0.587~0.804;声强特征在识别女性快感缺失型抑郁症中,敏感度70.0%,特异度58.1%,AUC为0.751,95%CI=0.554~0.769。结论梅尔倒谱系数、共振峰、声强和能量特征在女性快感缺失型抑郁症患者中可能存在特异性改变;其中梅尔倒谱系数对女性抑郁症患者快感缺失症状的识别准确率最高,有望成为女性快感缺失型抑郁症的客观评价指标。

Objective To explore the speech features of female patients with anhedonic depression and their recognition of pleasure deficient symptoms.Methods A total of 102 female depression patients who were hospitalized at Nanjing Brain Hospital from September 2020 to October 2021 were selected,including 62 anhedonic depression patients(anhedonic group)and 40 non-anhedonic depression patients(non-anhedonic group).A total of 50 female healthy controls were recruited during the same period.All participants were evaluated by the 17-item Hamilton depression scale(HAMD-17),Snaith-Hamilton pleasure scale(SHAPS),and the temporal experience of pleasure scale(TEPS),as well as voice acquisition.SPSS 23.0 software was used for data processing.Statistical analysis was conducted using one-way ANOVA,non-parametric tests,Logistic regression,and receiver operating characteristic curve.Results Compared with the non-anhedonic group,the anhedonic group showed significant changes in 15 voice features(all P<0.05),including Mel-frequency cepstral coefficients,formant frequencies,intensity,and energy features.Among these features,Mel-frequency cepstral coefficients exhibited the highest accuracy in identifying anhedonic depression,with sensitivity of 47.5%,specificity of 91.9%,area under curve(AUC)of 0.751,95%CI=0.686-0.866.Formant frequencies could identify female anhedonic depression,with a sensitivity of 90.0%,a specificity of 40.3%,an AUC of 0.647,and 95%CI=0.605-0.824.Energy features could identify anhedonic deficient depression,with a sensitivity of 60.0%,a specificity of 74.2%,an AUC of 0.679,and 95%CI=0.587-0.804.Intensity features could identify female anhedonic depression,with a sensitivity of 70.0%,a specificity of 58.1%,an AUC of 0.640,and 95%CI=0.554-0.769.Conclusion Mel-frequency cepstral coefficients,formant frequencies,intensity features,and energy features may have specific changes in female patients with anhedonic depression.The Mel-frequency cepstral coefficients has the highest recognition accuracy for anhedonic symptoms in female depression patients,and is expected to become an objective evaluation index for female anhedonic depression.