ORIGINAL ARTICLE
Year : 2023 | Volume
: 13 | Issue : 1 | Page : 40--48
Binary logistic regression modeling of voice impairment and voice assessment in iranian patients with nonlaryngeal head-and-neck cancers after chemoradiation therapy: Objective and subjective voice evaluation
Sakineh Bagherzadeh1, Daryoush Shahbazi-Gahrouei1, Farhad Torabinezhad2, Seied Rabi Mehdi Mahdavi3, Pedram Fadavi4, Soraya Salmanian4,
1 Department of Medical Physics, School of Medicine, Isfahan University of Medical Science, Isfahan, Iran
2 Department of Speech and Language Pathology, School of Rehabilitation Sciences, Iran University of Medical Sciences, Tehran, Iran
3 Department of Medical Physics, School of Medicine, Iran University of Medical Sciences, Tehran, Iran
4 Department of Radiotherapy, School of Medicine, Iran University of Medical Sciences, Tehran, Iran
Correspondence Address:
Daryoush Shahbazi-Gahrouei
Department of Medical Physics, School of Medicine, Isfahan University of Medical Science, Isfahan
Iran
Abstract
Background: Laryngeal damages after chemoradiation therapy (RT) in nonlaryngeal head-and-neck cancers (HNCs) can cause voice disorders and finally reduce the patient's quality of life (QOL). The aim of this study was to evaluate voice and predict laryngeal damages using statistical binary logistic regression (BLR) models in patients with nonlaryngeal HNCs. Methods: This cross-section experimental study was performed on seventy patients (46 males, 24 females) with an average age of 50.43 ± 16.54 years, with nonlaryngeal HNCs and eighty individuals with assumed normal voices. Subjective and objective voice assessment was carried out in three stages including before, at the end, and 6 months after treatment. Eventually, the Enter method of the BLR was used to measure the odds ratio of independent variables. Results: In objective evaluation, the acoustic parameters except for F0 increased significantly (P < 0.001) at the end treatment stage and decreased 6 months after treatment. The same trend can be seen in the subjective evaluations, whereas none of the values returned to pretreatment levels. Statistical models of BLR showed that chemotherapy (P < 0.05), mean laryngeal dose (P < 0.05), V50 Gy (P = 0.002), and gender (P = 0.008) had the greatest effect on incidence laryngeal damages. The model based on acoustic analysis had the highest percentage accuracy of 84.3%, sensitivity of 87.2%, and the area under the curve of 0.927. Conclusions: Voice evaluation and the use of BLR models to determine important factors were the optimum methods to reduce laryngeal damages and maintain the patient's QOL.
How to cite this article:
Bagherzadeh S, Shahbazi-Gahrouei D, Torabinezhad F, Mahdavi SR, Fadavi P, Salmanian S. Binary logistic regression modeling of voice impairment and voice assessment in iranian patients with nonlaryngeal head-and-neck cancers after chemoradiation therapy: Objective and subjective voice evaluation.J Med Signals Sens 2023;13:40-48
|
How to cite this URL:
Bagherzadeh S, Shahbazi-Gahrouei D, Torabinezhad F, Mahdavi SR, Fadavi P, Salmanian S. Binary logistic regression modeling of voice impairment and voice assessment in iranian patients with nonlaryngeal head-and-neck cancers after chemoradiation therapy: Objective and subjective voice evaluation. J Med Signals Sens [serial online] 2023 [cited 2023 Jun 5 ];13:40-48
Available from: https://www.jmssjournal.net/text.asp?2023/13/1/40/372559 |
Full Text
Introduction
Head-and-neck cancers (HNCs) are a group of cancers that involve several anatomical sites in the head-and-neck area.[1] Radiation therapy (RT) is a major method in the treatment of HNCs that can be prescribed with or without chemotherapy.[2] One of the problems of RT is to reach the damage to healthy organs near the tumor.[3] In most HNCs, the normal larynx is located in radiation fields due to its placement in the neck area and receiving high radiation doses.[4] Since the voice originates from the larynx, any vocal problems are a sign of laryngeal damages and dysfunction of the vocal cords. The voice is a multidimensional phenomenon and all quantitative and qualitative aspects of vocal problems should be evaluated.[3] Acoustical analysis of voice signals,[4],[5] self-assessment by questionnaires,[6],[7],[8] and perceptual assessment are well-known methods of voice examination.[9],[10] Several studies have used either of these methods to evaluate vocal problems.[10],[11],[12],[13],[14] However, this is the first study that was performed for Persian speakers. As the vowel system is different in English and Persia,[15] it was necessary to evaluate vocal problems and laryngeal edema in Iranian patients with nonlaryngeal HNCs.
Laryngeal damages and vocal problems are among the most important complications to be understood from chemo-RT for HNCs.[16] One of the important issues in the group of heterogeneous patients with different HNCs is the difference in the type of cancer, the type of treatment (chemoradiation or only radiation), different values of dosimetric parameters, and the individual conditions of each patient such as age and gender. As a result, it can mean that there is a different likelihood of complication or outcome for each patient.
Considering the mentioned conditions, the two issues of the effect of each factor in incidence complication as well as the difference or similarity in the results of each evaluation approach need further investigation.
Identifying factors that influence damages plays an effective role in disease prevention.[17] The use of appropriate statistical models and accurate estimation methods in conjunction with clinical diagnostics can be effective in determining these factors appropriately.[18],[19] Binary logistic regression (BLR) is a statistical technique to analyze the relation of a dependent variable to one or more independent variables in the research plans.[20],[21],[22],[23],[24] In none of the previous studies so far, the results of different evaluation methods and the effect of each factor in the incidence of laryngeal damages have not been investigated. Accordingly, the aim of the present study was to assess vocal changes and laryngeal dysfunction, especially laryngeal edema through objective and subjective evaluation of voice, as well as develop statistical models to improve disease detection by investigating effective factors in laryngeal dysfunction. For this purpose, BLR models were used to compare the performance of each of the assessment methods in classifying patients in the damaged and normal groups.
Materials and Methods
Seventy patients including 46 (65.71%) males and 24 (34.28%) females with a mean age of 50.43 ± 16.54 years, with various nonlaryngeal HNCs, were recorded in this cross-section experimental study. Descriptive analysis and related details about the types of malignancy and patient's conditions are given in [Table 1]. This experimental study was conducted in Haft Tir Hospital (Tehran, Iran) between December 2018 and September 2020. The present study was conducted following the approval of the Ethical Committee of Isfahan University of Medical Sciences (Isfahan, Iran, with ethical code number of IR. MUI. MED. REC.1398.041). All patients and the control group signed the consent form. The criteria for the patient's admission include no history of laryngeal disease and having a normal voice.{Table 1}
Chemoradiation therapy and dosimetry
Three-dimensional conformal RT (3D-CRT) was performed using a Siemens PRIMUS medical linear accelerator (Siemens AG, Erlangen, Germany) with MV energy located in Haft Tir, RT center. Before starting treatment, all patients were immobilized in the supine position for computed tomography (CT) simulation (Siemens Healthcare, Forchheim, Germany) and a four-point mask was used for each patient to repeat the position during treatment. CT images were obtained with 3-mm slice thickness. The CT images were converted into the Digital Imaging and Communications in Medicine format and transferred to the ISOgray treatment planning system (TPS) (DOSIsoft, French). The gross tumor volume, clinical target volume, and planning target volume as well as larynx as an organ at risk (OAR) were contoured on each slice of the CT images.[25] The doses of RT depending on the type of tumor ranged from 46 to 70 Gy for neck lymphoma and nasopharynx cases, respectively. All patients were treated five times a week, consecutively at a dose of fraction 1.8–2 Gy per day. In addition to RT, a number of patients (28 males, 15 females) had cisplatin-based chemotherapy, once a week. In order to calculate the mean, minimum, and maximum doses and V50 Gy in 27% of the larynx volume, a dose–volume histogram (DVH) was extracted from the TPS.
Voice evaluation
Three methods of voice evaluation including acoustic analysis, self-assessment use of Persian Voice Handicap Index (VHI) questionnaire, and perceptual evaluation were used to diagnose voice disorders and laryngeal damages in patients with nonlaryngeal HNCs. Data from eighty normal individuals were obtained as a control group. The eligible criteria for the individual patient were the larynx had to be uninvolved, the larynx had been in the radiation field, patients' voices should be normal, and all patients were followed at least once. People should not have any difficulty swallowing. In all evaluation methods, if the pretreatment data for patients were in line with the aforementioned control group's data, patients were entered in the relevant experiments.
Objective assessment
In a quiet room, the voices of all patients were recorded. All patients sustained the vowel/a/twice in the most habitual situations for 5 s, then counted the numbers from 1 to 10, and read the standard text “Pedarbozorg.” This text was developed to assess voice samples in Iranian patients with vocal dysfunctions.[26] The participants were instructed to produce an/a:/as long as possible after taking a maximal inhalation at a comfortable pitch and at their habitual loudness. Maximum phonation time (MPT) was expressed in seconds.
The recordings were conducted with a sampling frequency of 44.1 kHz and transferred to the computer system. Three seconds of vowel/a/and the whole time of other recorded voices were selected and analyzed using PRAAT software (version, 6.0.25) (Boersma,& Weenink, 2015) software (Finger,Cielo, Schwarz, 2015). The voices were recorded with the Zoom (H5, Japan) recorder.[27]
Subjective assessment
Patients were given a Persian version of the VHI questionnaire. This questionnaire has 30 questions in three subgroups; emotional, functional, and physical. Each subgroup has 10 questions and each question is rated from 0 to 4 scores. The total score of the questionnaire is 120. In the Persian VHI questionnaire, the cutoff point for the presence of voice impairments is 14.5.[28]
Perceptual evaluation
All patients read the standard text of the “Pedarbozorg,” again inconsistent conditions. Perceptual evaluation of vocal disorders was performed using of GRBAS scale. This scale consisted of G (grade), R (rough), B (breathy), A (asthenia), and S (straining). The mean overall voice severity parameter “G,” from the GRBAS ranged from G = 0–3 (0 = normal, 1 = mild, 2 = moderate, and 3 = severe) considered for each patient.[27] Two experienced speech therapists, completely unaware of the patient's condition evaluated any voice samples separately. All the mentioned tests were repeated in three times points including before, end, and 6 months after treatment. Sustaining the vowel was used for acoustic analysis of voice signals, numbers were counted to obtain habitual frequency, and “Pedarbozorg” text was used for perceptual evaluation. A total of 840 sound samples were obtained from patients and 240 samples from the control group.
Data analysis methodology
Binary logistic regression model
Logistic regression (LR) was introduced in the late 1960s and became popular among all researchers, especially health researchers.
Regression analysis presents the association between an outcome (dependent) variable and one or more predicted (independent) variables. BLR is used when the outcome variable had two categories.[22],[27] In this study, laryngeal damages were considered as dependent variables (presence = 1, absent = 0) and chemotherapy, mean, maximum, and minimum doses in the larynx, V50 Gy in the volume of 27% or higher of the larynx, smoking status, age, and gender were considered as independent variables. At the first, the collinearity statistical test was carried out to investigate the collinearity between the predictor variables. For all predictor variables, there were no collinearity trends, (tolerance >0.4).
For all voice evaluation methods, univariate and multivariate BLR models were computed. In the univariate analysis, the results from the LR and the Pearson Chi-square test were calculated and compared.
Variables with (P < 0.1) in the univariate analysis were entered into the multivariate analysis. The Enter method of BLR was used to measure the odds ratio (OR) of each variable in this research at the P level.
The results for three BLR models were evaluated in terms of performance with receiver operating characteristics curve and the area under the curve. One of the best ways to validate the damage was to perform a laryngoscopy, but this was avoided due to the damage caused by the RT and the problems for the patients.
Statistical analysis
The normality of all data was evaluated with the Kolmogorov–Smirnov test. The data for the control group and patients before treatment presented a normal distribution (P > 0.05). Independent samples t-test was used to compare the mean values between one standard deviation (mean ± SD) of acoustic parameters and questionnaire subgroups scores in two groups. Due to the abnormal distribution (P < 0.05) of patient's data at three times points (before, end, and 6 months after treatment), the nonparametric Friedman test was used to compare over time. All analyses were performed with the use of SPSS software (IBM Crop. 2019. IBM SPSS, version 26.0, IBM Corp., Armonk, N.Y., USA), and the significance level of P < 0.05 was considered in this study.
Results
The values of important acoustic parameters including jitter (%) and shimmer (%) (period-to-period irregularities in frequency and in amplitude), respectively, main fundamental frequency (F0), habitual frequency pitch, harmonic-to-noise ratio (HNR), and noise-to-harmonic ratio (NHR) were extracted using of PRAAT software. The results of the independent samples t-test related to objective and subjective evaluations for the control group and patients before treatment are shown in [Table 2]. There was no significant difference (P > 0.05) in the mean ± SD values between the two groups in both evaluation methods. In contrast, the results of the Friedman test for acoustic analysis in the patient's group are shown in [Figure 1], Part A. Results show significant changes in the mean ranks for all acoustic parameters (P < 0.001), except main fundamental frequency (male P = 0.316, female P = 0.214), frequency range (male P = 0.221, female P = 0.311), and MPT (male P = 0.411, female P = 0.219). Acoustic parameters increased during the treatment and decreased 6 months after treatment. Changes for HNR were in the opposite direction of the mentioned parameters. The results of the Persian VHI questionnaire and the perceptual assessment are given in Sections B and C of [Figure 1]. For the Persian VHI questionnaire, the mean rank of the values in all three subgroups and the total group increased significantly (P < 0.001) during the treatment and decreased in 6 months after treatment but did not reach the base level. Interrater agreement for GRBAS rating between two experienced experts was measured using the intraclass correlation coefficient, and the excellent result was (0.921). Perceptual evaluation of voice quality was measured by GRBAS grade. Friedman test results and frequency show significant changes (P < 0.001) between each of grade values at any given time.{Figure 1}{Table 2}
Binary logistic regression models
Acoustic analysis
According to the acoustic analysis 6 months after treatment, 31 (44.28%) patients were classified in the damaged group. The results of the univariate analysis using the LR method (likelihood ratio) and the Pearson Chi-square coefficient were virtually the same for all variables. The omnibus tests demonstrated that the resulting model is significant (Chi-square = 48.330, df = 6, P < 0.001 and Cos and Snell R2 = 0.499, Nagelkerke R2 = 0.668).
In multivariate analysis, the variables of chemotherapy, mean laryngeal dose, and V50 Gy showed a significant level (P < 0.05). The BLR result showed a true classification rate of 87.1%. In other words, the regression model classified 27 out of 31 cases exactly, as expected in [Table 3].{Table 3}
Subjective assessment
Persian Voice Handicap Index questionnaire
Based on the cutoff value for vocal problems in the Persian VHI questionnaire, 42 (60%) patients with a score higher than 15 were classified as damaged.
The results of omnibus test for this model such as Chi-square and df were 18.430 and 4, respectively (P < 0.001). Independent variables can predict 20%–28% of the variance of laryngeal damage (Cos and Snell R2 = 0.209, Nagelkerke R2 = 0.283). Multivariate analysis showed that chemotherapy and mean laryngeal dose are the important factors with P < 0.05. The true classification rate was 83.3%; in other words, the regression model classified 35 out of 42 cases correctly; [Table 4].{Table 4}
Perceptual assessment
Based on perceptual evaluation, 37 (52.85%) patients had high G-score >1.5 and were classified in the damaged group. In multivariate analysis, the variables of gender and mean laryngeal dose were the influential variables with a significant level of P < 0.05. Results from omnibus tests demonstrated that model is significant (Chi-square = 19.773, df = 4, P < 0.001) and can predict 24%–32% of the variance of laryngeal damage (Cos and Snell R2 = 0.246, Nagelkerke R2 = 0.328). The true classification rate was 81.1%, and 29 of the 37 patients were classified correctly [Table 5]. The comparison results of the three models are shown in [Table 6]. ROC diagrams for each model are also shown in [Figure 2].{Figure 2}{Table 5}{Table 6}
Discussion
Two tasks were performed in this experimental study: the first the assessment of voice as an important factor in the diagnosis of laryngeal edema and the second predicting the likelihood of damage based on various factors and use of statistical BLR models.
In this study, compared to the control group and pretreatment data, the values of the acoustic parameters increased at the end of treatment. Hoarseness, breathiness, and roughness are common indicators of change in voice signals. In the presence of vocal problems, these signs increase. The increase in values of perturbation parameters (jitter, shimmer, and NHR) indicates hoarseness of the voice. A decrease in HNR indicates the presence of breathiness, and roughness is related to the increased values of the main fundamental frequency and habitual frequency.[27],[28]
Dosimetric predictors for laryngeal edema after radiotherapy showed that at mean doses higher than 43.5–44 Gy, edema of Grade II or higher occurs.[29]
Based on DVH data, 39 (55.71%) patients had a mean laryngeal dose of higher than 44 Gy. In these patients, the values of the perturbation parameters increased which was an indication of edema in the vocal cords.
Six months after treatment by reducing the side effects of treatment on the larynx, the values of acoustic parameters were reduced but do not return to baseline level. HNR values increased at the end of treatment.
In general, there are vocal problems that none of the instrumental and visual methods can recognize and only the patient can understand intrinsically which needs to be noted. The Persian VHI questionnaire is a valuable tool for evaluating these problems and the patient's quality of life (QOL).[15],[30],[31]
An increase in VHI questionnaire scores is consistent with the increase in the rate of vocal problems and reduced voice-related QOL.[7] In the final evaluation, 42 (60%) of patients believed that their voice was different than before and complained of voice problems such as breathiness and hoarseness. In the present study, most of the patients had tumors above of hyoid bone (oropharyngeal, base of tongue, tonsil, and nasopharyngeal). RT affects the speech of patients with tumors above the hyoid bone more than voice.[8]
The patient's speech was examined by counting the numbers from 1 to 10, and reading the standard text of the “Pedarbozorg”. The results of the latest evaluation by two experienced speech therapists based on the GRBAS scale showed that 37 (53%) of patients had speech problems. The results obtained from the subjective evaluation revealed that the majority of patients with higher VHI scores also had a larger G value in perceptual evaluation. These results were in agreement with the results of previous studies.[12],[30]
One of the problems of patients following the RT is the complications caused by treatment and the consequent decrease in QOL. Hence, anticipating the incidence of damage for each patient will be valuable. A predictive model based on LR for radiation pneumonitis in patients receiving modern chemo-RT for the treatment of locally advanced nonsmall-cell lung carcinoma (NSCLC) was developed by Palma. They concluded that the type of chemotherapy regimen, dosimetric parameters, and patient age were the important factors in the incidence of NSCLC.[32] Langendijk et al. found that swallowing dysfunction can reduce the patient's QOL even more than xerostomia. Therefore, they tried to develop a model that can be easily used in clinical applications and predict dysphagia in patients under RT. Multivariate analyses of LR showed that T classification (T1–T4), neck irradiating (ipsilateral or both), weight loss, primary tumor site, and treatment modality such as conventional, accelerated, and concomitant chemotherapy were the important factors.[33]
In this study, for three models, multivariate analysis according to the Enter method of BLR showed that mean doses of larynx and chemotherapy were the most influential variables and had higher OR. In the first model, based on acoustic analysis, in addition to the mean dose of larynx and chemotherapy, V50 Gy had a significant level (P < 0.05, OR = 14.292). According to quantitative analyses of normal tissue effects in the clinic criteria, a dose of 50 Gy at 27% or larger volume of the larynx causes damage.[34]
In the third model, based on GRBAS, the gender factor had a significant level (P < 0.05, OR = 5.656). Investigators indicated that gender cannot be an important factor in the occurrence of damage.[23] The unequal number of male and female patients, heterogeneity and diversity in types of HNCs, and evaluation based on individual judgment may be the reasons for the difference between these results with other studies.
The results of [Table 6] and the ROC diagrams in [Figure 2] show that the model obtained based on acoustic analysis has the highest sensitivity, accuracy, and specificity, and the models based on subjective evaluations have provided almost identical results. Wide-field radiation in the 3D-CRT increased the OARs dose as well as larynx. The use of Intensity modulation RT and TomoTherapy techniques can reduce the dose of OARs and side effects of treatment and maintain the patient's QOL at an acceptable level. Accordingly, these results suggest the use of modern RT methods, increased sample size and long-term follow-up can lead to better outcomes.
Conclusions
Chemo-RT in patients with nonlaryngeal tumor HNCs can damage the larynx, thus causing vocal problems. Acoustic analysis, perceptual evaluation, and patient self-assessment are the best assessment methods due to their cheapness, noninvasiveness, high repeatability, and quantitative and qualitative evaluation of voice. Based on BLR analysis, the factors such as chemotherapy, mean dose of larynx, V50 Gy, and gender had the greatest effect on laryngeal damage and the regression model based on acoustic analysis had the best performance in classifying patients in affected and normal groups.
Financial support and sponsorship
This work is a part of PhD thesis which is financially supported (grant no: 397780 with ethical code no: IR. MUI. MED. REC.1398.041) by the Isfahan University of Medical Sciences, Isfahan, Iran.
Conflicts of interest
There are no conflicts of interest.
References
| 1 | Mahmood H, Shaban M, Rajpoot N, Khurram SA. Artificial Intelligence-based methods in head and neck cancer diagnosis: An overview. Br J Cancer 2021;124:1934-40. |
| 2 | Baskar R, Lee KA, Yeo R, Yeoh KW. Cancer and radiation therapy: Current advances and future directions. Int J Med Sci 2012;9:193-9. |
| 3 | Lazarus CL. Effects of chemoradiotherapy on voice and swallowing. Curr Opin Otolaryngol Head Neck Surg 2009;17:172-8. |
| 4 | Brockmann M, Drinnan MJ, Storck C, Carding PN. Reliable jitter and shimmer measurements in voice clinics: The relevance of vowel, gender, vocal intensity, and fundamental frequency effects in a typical clinical task. J Voice 2011;25:44-53. |
| 5 | Vaziri G, Almasganj F, Behroozmand R. Pathological assessment of patients' speech signals using nonlinear dynamical analysis. Comput Biol Med 2010;40:54-63. |
| 6 | Kraaijenga SA, Oskam IM, van Son RJ, Hamming-Vrieze O, Hilgers FJ, van den Brekel MW, et al. Assessment of voice, speech, and related quality of life in advanced head and neck cancer patients 10-years+after chemoradiotherapy. Oral Oncol 2016;55:24-30. |
| 7 | Schindler A, Mozzanica F, Vedrody M, Maruzzi P, Ottaviani F. Correlation between the voice handicap index and voice measurements in four groups of patients with dysphonia. Otolaryngol Head Neck Surg 2009;141:762-9. |
| 8 | van der Molen L, van Rossum MA, Jacobi I, van Son RJ, Smeele LE, Rasch CR, et al. Pre- and posttreatment voice and speech outcomes in patients with advanced head and neck cancer treated with chemoradiotherapy: Expert listeners' and patient's perception. J Voice 2012;26:664.e25-33. |
| 9 | Nemr K, Simões-Zenari M, Cordeiro GF, Tsuji D, Ogawa AI, Ubrig MT, et al. GRBAS and Cape-V scales: High reliability and consensus when applied at different times. J Voice 2012;26:812.e17-22. |
| 10 | Hamdan AL, Geara F, Rameh C, Husseini ST, Eid T, Fuleihan N. Vocal changes following radiotherapy to the head and neck for non-laryngeal tumors. Eur Arch Otorhinolaryngol 2009;266:1435-9. |
| 11 | Bibby JR, Cotton SM, Perry A, Corry JF. Voice outcomes after radiotherapy treatment for early glottic cancer: Assessment using multidimensional tools. Head Neck 2008;30:600-10. |
| 12 | Davies-Husband C, Murphy J, Kelly C, Drinnan M, Paleri V. Extreme long-term voice outcomes after concurrent chemoradiotherapy for advanced non-laryngeal head and neck cancer: Eight-year post-treatment analysis. Clin Otolaryngol 2018;43:1494-9. |
| 13 | Paleri V, Carding P, Chatterjee S, Kelly C, Wilson JA, Welch A, et al. Voice outcomes after concurrent chemoradiotherapy for advanced nonlaryngeal head and neck cancer: A prospective study. Head Neck 2012;34:1747-52. |
| 14 | Radhakrishna N, Yamini BK, Kadam AS, Shivashankar N, Vishwanathan C, Javarappa R. Acoustic analysis of voice in nonlaryngeal head and neck cancer patients post chemoradiotherapy. J Cancer Res Ther 2017;13:113-7. |
| 15 | Dehqan A, Yadegari F, Scherer RC, Dabirmoghadam P. Correlation of VHI-30 to acoustic measurements across three common voice disorders. J Voice 2017;31:34-40. |
| 16 | Heijnen B, Speyer R, Kertscher B, Cordier R, Koetsenruijter K, Swan K, et al. Dysphagia, speech, voice, and trismus following radiotherapy and/or chemotherapy in patients with head and neck carcinoma: review of the literature. BioMed research international 2016;2016:1-25. |
| 17 | Prasanna PG, Stone HB, Wong RS, Capala J, Bernhard EJ, Vikram B, et al. Normal tissue protection for improving radiotherapy: Where are the Gaps? Transl Cancer Res 2012;1:35-48. |
| 18 | Ellies-Oury MP, Chavent M, Conanec A, Bonnet M, Picard B, Saracco J. Statistical model choice including variable selection based on variable importance: A relevant way for biomarkers selection to predict meat tenderness. Sci Rep 2019;9:10014. |
| 19 | Henley SS, Golden RM, Kashner TM. Statistical modeling methods: Challenges and strategies. Biostat Epidemiol 2020;4:105-39. |
| 20 | Bagley SC, White H, Golomb BA. Logistic regression in the medical literature: Standards for use and reporting, with particular attention to one medical domain. J Clin Epidemiol 2001;54:979-85. |
| 21 | Morimoto M, Bijl HP, VAN DER Schaaf A, Xu CJ, Steenbakkers RJ, Chouvalova O, et al. Development of normal tissue complication probability model for trismus in head and neck cancer patients treated with radiotherapy: The role of dosimetric and clinical factors. Anticancer Res 2019;39:6787-98. |
| 22 | Shanbehzadeh M, Nopour R. Determination of the most important diagnostic criteria for COVID-19: A step forward to design an intelligent clinical decision support system. J Adv Med Biomed Res 2021;29:176-82. |
| 23 | Singh V, Dwivedi SN, Deo SV. Ordinal logistic regression model describing factors associated with extent of nodal involvement in oral cancer patients and its prospective validation. BMC Med Res Methodol 2020;20:95. |
| 24 | Tikka T, Kavanagh K, Lowit A, Jiafeng P, Burns H, Nixon IJ, et al. Head and neck cancer risk calculator (HaNC-RC)–V. 2. Adjustments and addition of symptoms and social history factors. Clin Otolaryngol 2020;45:380-8. |
| 25 | Choi M, Refaat T, Lester MS, Bacchus I, Rademaker AW, Mittal BB. Development of a standardized method for contouring the larynx and its substructures. Radiat Oncol 2014;9:285. |
| 26 | Aghajanzadeh M, Ghorbani A, Torabinezhad F, Reza Keyhani M. Comparing of two Farsi passages in laboratory evaluation of mean and fundamental frequency variations range. Audiology. 2012;21:62-68. |
| 27 | Teixeira JP, Fernandes PO. Acoustic analysis of vocal dysphonia. Procedia Comput Sci 2015;64:466-73. |
| 28 | Teixeira JP, Oliveira C, Lopes C. Vocal acoustic analysis-jitter, shimmer and hnr parameters. Procedia Technol 2013;9:1112-22. |
| 29 | Sanguineti G, Adapala P, Endres EJ, Brack C, Fiorino C, Sormani MP, et al. Dosimetric predictors of laryngeal edema. Int J Radiat Oncol Biol Phys 2007;68:741-9. |
| 30 | Niebudek-Bogusz E, Woznicka E, Zamyslowska-Szmytke E, Sliwinska-Kowalska M. Correlation between acoustic parameters and Voice Handicap Index in dysphonic teachers. Folia Phoniatr Logop 2010;62:55-60. |
| 31 | Webb AL, Carding PN, Deary IJ, MacKenzie K, Steen N, Wilson JA. The reliability of three perceptual evaluation scales for dysphonia. Eur Arch Otorhinolaryngol 2004;261:429-34. |
| 32 | Palma DA, Senan S, Tsujino K, Barriger RB, Rengan R, Moreno M, et al. Predicting radiation pneumonitis after chemoradiation therapy for lung cancer: an international individual patient data meta-analysis. International Journal of Radiation Oncology* Biology* Physics 2013;85:444-50. |
| 33 | Langendijk JA, Doornaert P, Rietveld DH, Verdonck-de Leeuw IM, Leemans CR, Slotman BJ. A predictive model for swallowing dysfunction after curative radiotherapy in head and neck cancer. Radiotherapy and Oncology 2009;90:189-95. |
| 34 | Rancati T, Schwarz M, Allen AM, Feng F, Popovtzer A, Mittal B, et al. Radiation dose-volume effects in the larynx and pharynx. Int J Radiat Oncol Biol Phys 2010;76:S64-9. |
|
