基于MFCC-DTW距离的甲状腺术后嗓音恢复监测研究OA

Objective To quantitatively assess voice recovery following thyroid surgery with the Mel-frequency cepstral coefficient(MFCC)-based dynamic time warping(DTW)distance(MFCC-DTW distance),so as to solve the problems of postoperative voice recovery monitoring in lacking objective evaluation methods.Methods Voice signals were collected from 54 patients undergoing thyroid nodule resection 1 day before surgery and 12,24,48 hours and 6 months after surgery,and the scores for the Grade,Roughness,Breathiness,Asthenia and Strain(GRBAS)were recorded simultaneously.The spectral characteristics and MFCC features of preoperative and postoperative voice signals were analyzed.The DTW algorithm was then used to calculate the DTW distance between preoperative and postoperative MFCC features,thereby enabling monitoring and evaluation of vocal function recovery.The consistency between MFCC-DTW distance and GRBAS scores was further explored.Results Changes in the spectral characteristics and MFCC features of patients' voice signals were highly consistent with their recovery status.The mean MFCC-DTW distances between the postoperative voices of the 54 patients at 12,24,and 48 hours and at 6 months and their preoperative voices were 246,204,190 and 85,respectively,showing a gradual decrease as recovery progressed(P＜0.05).The Spearman correlation coefficients between the MFCC-DTW distance and the GRBAS scores at 12,24 and 48 hours postoperatively were 0.908,0.903 and 0.910,respectively(all P＜0.01).Most the patients regained largely their vocal function 6 months after surgery,with GRBAS scores concentrated at 0 or close to 0;the Spearman correlation coefficient between the MFCC-DTW distance and the GRBAS score was 0.642.Conclusion The proposed method effectively quantifies the recovery progress of voice function after thyroid surgery,and facilitates convenient and reliable follow-up evaluation of postoperative voice recovery.[Chinese Medical Equipment Journal,2026,47(6):20-28]