Voice changes may be the earliest signs in laryngeal cancer. We investigated whether automated voice signal analysis can be used to distinguish patients with laryngeal cancer from healthy subjects. We extracted features using the software package for speech analysis in phonetics (PRAAT) and calculated the Mel-frequency cepstral coefficients (MFCCs) from voice samples of a vowel sound of /a:/. The proposed method was tested with six algorithms: support vector machine (SVM), extreme gradient boosting (XGBoost), light gradient boosted machine (LGBM), artificial neural network (ANN), one-dimensional convolutional neural network (1D-CNN) and two-dimensional convolutional neural network (2D-CNN). Their performances were evaluated in terms of accuracy, sensitivity, and specificity. The result was compared with human performance. A total of four volunteers, two of whom were trained laryngologists, rated the same files. The 1D-CNN showed the highest accuracy of 85% and sensitivity and sensitivity and specificity levels of 78% and 93%. The two laryngologists achieved accuracy of 69.9% but sensitivity levels of 44%. Automated analysis of voice signals could differentiate subjects with laryngeal cancer from those of healthy subjects with higher diagnostic properties than those performed by the four volunteers.

Background and objectives: This study aimed to determine the cut-off values of the following three respiratory pressure meters; the voluntary peak cough flow (PCF), maximal expiratory pressure (MEP) and maximal inspiratory pressure (MIP); associated with post-stroke dysphagia and assess which of these parameters show good diagnostic properties associated with post-stroke dysphagia. Materials and Methods: Retrospective analysis of a prospectively maintained database. Records of patients with first-ever diagnosed dysphagia attributable to cerebrovascular disease, who had performed spirometry measurements for the PCF, MIP and MEP. Results: From a total of 237 stroke patients, 163 patients were diagnosed with dysphagia. Those with dysphagia had significantly lower PCF values than those without dysphagia (116.3 ± 75.3 vs. 219.4 ± 91.8 L/min, p < 0.001). In addition, the former group also had lower MIP (30.5 ± 24.7 vs. 41.6 ± 25.7 cmH₂O, p = 0.0002) and MEP (41.0 ± 27.9 vs. 62.8 ± 32.3 cmH₂O, p < 0.001) values than the latter group. The receiver operating characteristic curve analysis showed that the PCF cut-off value of 151 L/min (area under the receiver operating characteristic curve [AUC] 0.81; sensitivity 72%; specificity 78.8%) was associated with post-stroke dysphagia. The optimum MEP and MIP cut-off were 38 cmH₂O (AUC 0.70, sensitivity 58%; specificity 77.7%) and 20 cmH₂O (AUC 0.65, sensitivity 49%; specificity 84%). PCF showed the highest AUC results. Results from the univariate analysis indicated that PCF values of ≤151 L/min increased risk of dysphagia by 9.51-fold (4.96–18.23). Multivariable analysis showed that after controlling of other clinical factor, the PCFs at this cut-off value still showed increased risk of by 4.19 (2.02–83.69) but this was not observed with the MIPs or MEPs. Conclusions: Our study has provided cut-off values that are associated with increased risk of dysphagia. Among the three parameters, PCF showed increased association with post-stroke dysphagia.