369例口腔癌患者汉语普通话语音数据库的建立与初步评价

doi:10.19438/j.cjoms.2022.02.009

摘要/Abstract

摘要： 目的:收集口腔癌患者手术前、后语音样本,建立口腔癌患者汉语普通话语音数据库,为口腔癌的临床诊治和康复提供数据平台。方法:收集在中山大学附属口腔医院口腔颌面外科就诊及术后定期复诊的口腔癌患者,利用多种语音测试工具对患者进行语音评估,在安静环境下录音,纳入非恶性肿瘤病变患者和健康人作为对照组。所有音频样本经统一预处理、分割、标注、匿名化处理,生成标准化语音数据集。结果:2017年7月—2021年4月纳入481位独立发音人,男274例（57.0%）,女207例（43.0%）,平均年龄（46.98±16.34）岁。口腔癌患者369例（76.7%）,健康人79例（16.4%）,非恶性肿瘤患者33例（6.9%）。口腔癌患者中,病变位于舌及口底258例,术前进入评估队列202例,T1、T2、T3、T4分别为35例（17.3%）、68例（33.7%）、41例（20.3%）、58例（28.7%）。所有发音人的中位随访（评估）时间为术后219 d（IQR：87.5~587 d）,共计1 100人次。共得到73 008份独立语料音频,所涉及的语料种类包括元音（6,1.30%）、下颌轮替运动音节（7,1.51%）、单字（238,51.63%）、词语（169,36.66%）、句子（38,8.24%）共461种。结论:本研究建立了国内外首个以口腔癌为主要病种的语音数据库,丰富了口腔癌临床诊治的功能学维度,为临床生物信息标志物的深入研究及个体化语言康复提供了重要的数据支持。

关键词: 数据库, 语音, 口腔癌, 普通话

Abstract: PURPOSE: To establish Mandarin speech database of Chinese oral cancer patients by collecting speech samples of patients with oral cancers before and after operation, so as to provide important data platform for clinical diagnosis, treatment and rehabilitation of oral cancer patients. METHODS: The patients from Department of Oral and Maxillofacial Surgery, Hospital of Stomatology, Sun Yat-Sen University were collected. A variety of speech corpus were applied to assess the participants and collect speech samples which were followed by uniform pre-processing, segmentation, annotation, and anonymization. A set of speech samples from relative health controls were also collected. RESULTS: A total of 481 individual speakers were enrolled from July 2017 to April 2021. The mean age at first assessment was 46.98±16.34 years. The participants consisted of 274 males(57.0%) and 207 females(43.0%). Of them, 369 patients with oral cancers（76.7%）, 79 healthy subjects(16.4%) and 33 subjects with non-malignant lesions(6.9%) were comprised. Among oral cancers, a sum of 258 patients had primary lesions located on the tongue and floor of the mouth, and 202 patients were recruited before surgery whose T classifications were 35 cases at T1(17.3%), 68 cases at T2(33.7%), 41 cases at T3(20.3%) and 58 cases at T4(28.7%). The median follow-up(assessment) time was 219 d(IQR: 87.5~587 d) after surgery with a sum of 1 100 appointments. A sum of 73 008 isolated audio samples were obtained from 461 kinds of corpus stimuli which consisted of vowels (6,1.30%), diadochokinesis syllables(7, 1.51%), single words (238, 51.63%), phrases (169, 36.66%) and sentences (38,8.24%). CONCLUSIONS: The present study established the first speech database whose speech samples were dominantly from oral cancers, which provided important data support for the in-depth study of clinical biomarkers and development of individualized speech rehabilitation.

Key words: Database, Speech, Oral cancer, Mandarin Chinese

中图分类号:

R739.8

肖育栋, 郭凯欣, 杨乐, 邓威, 曾滨, 张思恩, 劳小媚, 廖贵清, 梁玉洁. 369例口腔癌患者汉语普通话语音数据库的建立与初步评价[J]. 中国口腔颌面外科杂志, 2022, 20(2): 151-157.

XIAO Yu-dong, GUO Kai-xin, YANG Le, DENG Wei, ZENG Bin, ZHANG Si-en, LAO Xiao-mei, LIAO Gui-qing, LIANG Yu-jie. Preliminary establishment of Mandarin speech database of369Chinese oral cancer patients[J]. China J Oral Maxillofac Surg, 2022, 20(2): 151-157.

参考文献

[1] WHO-IARC. Globocan2020: Estimated cancer incidence, mortality and prevalence worldwide in 2020[EB/OL]. Globocan, 2020, https://gco.iarc.fr/today/online-analysis-map.
[2] Chen W, Xia C, Zheng R, et al.Disparities by province, age, and sex in site-specific cancer burden attributable to 23 potentially modifiable risk factors in china: a comparative risk assessment[J]. Lancet Glob Health, 2019, 7(2): e257-e269.
[3] Chi AC, Day TA, Neville BW.Oral cavity and oropharyngeal squamous cell carcinoma--an update[J]. CA Cancer J Clin, 2015, 65(5): 401-421.
[4] Colangelo LA, Logemann JA, Rademaker AW.Tumor size and pretreatment speech and swallowing in patients with resectable tumors[J]. Otolaryngol Neck Surg Surg, 2000, 122(5): 653-661.
[5] Kraaijenga SAC, Oskam IM, Van Son RJJH, et al.Assessment of voice, speech, and related quality of life in advanced head and neck cancer patients 10-years+ after chemoradiotherapy[J]. Oral Oncol, 2016, 55: 24-30.
[6] Laaksonen JP, Rieger J, Happonen RP, et al.Speech after radial forearm free flap reconstruction of the tongue: a longitudinal acoustic study of vowel and diphthong sounds[J]. Clin Linguist Phon, 2010, 24(1): 41-54.
[7] Gómez-Vilda P, Gómez-Rodellar A, Vicente JMF, et al.Neuromechanical modelling of articulatory movements from surface electromyography and speech formants[J]. Int J Neural Syst, 2019, 29(2): 1850039.
[8] Takatsu J, Hanai N, Suzuki H, et al.Phonologic and acoustic analysis of speech following glossectomy and the effect of rehabilitation on speech outcomes[J]. J Oral Maxillofac Surg, 2017, 75(7): 1530-1541.
[9] Gelfer MP, Bennett QE.Speaking fundamental frequency and vowel formant frequencies: effects on perception of gender[J]. J Voice, 2013, 27(5): 556-566.
[10] Jacewicz E, Fox RA.The effects of cross-generational and cross-dialectal variation on vowel identification and classification[J]. J Acoust Soc Am, 2012, 131(2): 1413-1433.
[11] Gómez-Vilda P, Mekyska J, Ferrández JM, et al.Parkinson disease detection from speech articulation neuromechanics[J]. Front Neuroinform, 2017, 11: 56-73.
[12] Jeancolas L, Mangone G, Corvol JC, et al.Comparison of telephone recordings and professional microphone recordings for early detection of parkinson’s disease, using mel-frequency cepstral coefficients with gaussian mixture models[C]. Graz: Interspeech 2019, 2019: 3033-3037.
[13] MacWhinney B, Fromm D, Forbes M, et al. Aphasiabank: methods for studying discourse[J]. Aphasiology, 2011, 25(11): 1286-1307.
[14] 马平川, 毛渤淳, 郭春丽, 等. 汉语普通话腭裂语音数据库的搭建与应用[J]. 华西口腔医学杂志, 2020, 38(2): 149-154.
[15] Harar P, Galaz Z, Alonso-Hernandez JB, et al.Towards robust voice pathology detection: investigation of supervised deep learning, gradient boosting, and anomaly detection approaches across four databases[J]. Neural Comput Appl, 2020, 32(20): 15747-15757.
[16] Serrano García L, Raman S, Hernáez Rioja I, et al.A spanish multispeaker database of esophageal speech[J]. Comput Speech Lang, 2021, 66: 101168.
[17] Alimuradov AK, Tychkov AY, Mezhina VA, et al.Development of natural emotional speech database for training automatic recognition systems of stressful emotions in human-robot interaction[C]. 2020 4th Scientific School on Dynamics of Complex Networks and their Application in Intellectual Robotics (DCNAIR), 2020: 11-16.
[18] Hadjadji I, Falek L, Lounnas K, et al.Enhancement of the interlocutor emotion recognition rate from non-professionals speakers in arabic database[C]. Morocco: The 2nd International Conference on Electronics, Control, Optimization and Computer Science 2020, 2020: 1-5.
[19] 黄昭鸣, 杜晓新. 言语障碍的评估与矫治[M]. 上海: 华东师范大学出版社, 2006: 118.
[20] Chen XR, Zhao B, Yin H.The charactristics of speech training for postoperation cleft palate patients who have marginal velopharyngeal insufficiency after palatoplasty[J]. Int J Stomatol, 2011, 38(3): 279-282.
[21] 王国民, 朱川, 袁文化, 等. 汉语语音清晰度测试字表的建立和临床应用研究[J]. 上海口腔医学, 1995, 4(3): 125-127.
[22] Jeancolas L, Petrovska-Delacrétaz D, Mangone G, et al.X-Vectors: new quantitative biomarkers for early parkinson’s disease detection from speech[J]. Front Neuroinform, 2021, 15: 578369.
[23] Nevler N, Ash S, Irwin DJ, et al.Validated automatic speech biomarkers in primary progressive aphasia[J]. Ann Clin Transl Neurol, 2019, 6(1): 4-14.
[24] Wang X, Tang M, Yang S, et al.Automatic hypernasality detection in cleft palate speech using CNN[J]. Circuits Syst Signal Process, 2019, 38(8): 3521-3547.
[25] Kim H, Jeon J, Han YJ, et al.Convolutional neural network classifies pathological voice change in laryngeal cancer with high accuracy[J]. J Clin Med, 2020, 9(11): 3415-3430.
[26] Xiao Y, Wang T, Deng W, et al.Data mining of an acoustic biomarker in tongue cancers and its clinical validation[J]. Cancer Med, 2021, 10(11): 3822-3835.
[27] Yi CR, Jeong WS, Oh TS, et al.Analysis of speech and functional outcomes in tongue reconstruction after hemiglossectomy[J]. J Reconstr Microsurg, 2020, 36(7): 507-513.
[28] Plamondon R.A kinematic theory of rapid human movements. Part I. Movement representation and generation[J]. Biol Cybern, 1995, 72(4): 295-307.
[29] Fischer A, Plamondon R.Signature verification based on the kinematic theory of rapid human movements[J]. IEEE Trans Human Machine Syst, 2017, 47(2): 169-180.
[30] Carmona-Duarte C, Ferrer MA, Plamondon R, et al.Sigma-lognormal modeling of speech[J]. Cognit Comput, 2021, 13(2): 488-503.
[31] Bouchard KE, Mesgarani N, Johnson K, et al.Functional organization of human sensorimotor cortex for speech articulation[J]. Nature, 2013, 495(7441): 327-332.