Comparison of supervised machine learning classification techniques in prediction of locoregional recurrences in early oral tongue cancer

Alabi R.O., Elmusrati M., Sawazaki-Calone I., Kowalski L.P., Haglund C., Coletta R.D., Mäkitie A.A., Salo T., Leivo I., Almangush A.

Estimation of risk of recurrence in early-stage oral tongue squamous cell carcinoma (OTSCC) remains a challenge in the field of head and neck oncology. We examined the use of artificial neural networks (ANNs) to predict recurrences in early-stage OTSCC. A Web-based tool available for public use was also developed. A feedforward neural network was trained for prediction of locoregional recurrences in early OTSCC. The trained network was used to evaluate several prognostic parameters (age, gender, T stage, WHO histologic grade, depth of invasion, tumor budding, worst pattern of invasion, perineural invasion, and lymphocytic host response). Our neural network model identified tumor budding and depth of invasion as the most important prognosticators to predict locoregional recurrence. The accuracy of the neural network was 92.7%, which was higher than that of the logistic regression model (86.5%). Our online tool provided 88.2% accuracy, 71.2% sensitivity, and 98.9% specificity. In conclusion, ANN seems to offer a unique decision-making support predicting recurrences and thus adding value for the management of early OTSCC. To the best of our knowledge, this is the first study that applied ANN for prediction of recurrence in early OTSCC and provided a Web-based tool.

Bur A.M., Holcomb A., Goodwin S., Woodroof J., Karadaghy O., Shnayder Y., Kakarala K., Brant J., Shew M.

To develop and validate an algorithm to predict occult nodal metastasis in clinically node negative oral cavity squamous cell carcinoma (OCSCC) using machine learning. To compare algorithm performance to a model based on tumor depth of invasion (DOI).Patients who underwent primary tumor extirpation and elective neck dissection from 2007 to 2013 for clinical T1-2N0 OCSCC were identified from the National Cancer Database (NCDB). Multiple machine learning algorithms were developed to predict pathologic nodal metastasis using clinicopathologic data from 782 patients.The algorithm was internally validated using test data from 654 patients in NCDB and was then externally validated using data from 71 patients treated at a single academic institution. Performance was measured using area under the receiver operating characteristic (ROC) curve (AUC). Machine learning and DOI model performance were compared using Delong's test for two correlated ROC curves.The best classification performance was achieved with a decision forest algorithm (AUC = 0.840). When applied to the single-institution data, the predictive performance of machine learning exceeded that of the DOI model (AUC = 0.657, p = 0.007). Compared to the DOI model, machine learning reduced the number of neck dissections recommended while simultaneously improving sensitivity and specificity.Machine learning improves prediction of pathologic nodal metastasis in patients with clinical T1-2N0 OCSCC compared to methods based on DOI. Improved predictive algorithms are needed to ensure that patients with occult nodal disease are adequately treated while avoiding the cost and morbidity of neck dissection in patients without pathologic nodal disease.

Yamakawa N., Kirita T., Umeda M., Yanamoto S., Ota Y., Otsuru M., Okura M., Kurita H., Yamada S., Hasegawa T., Aikawa T., Komori T., Ueda M.

Background and objectives Some patients with early-stage oral cancer have a poor prognosis owing to the delayed neck metastasis (DNM). Tumor budding is reportedly a promising prognostic marker in many cancers. Moreover, the tissue surrounding a tumor is also considered to play a prognostic role. In this study, we evaluated whether tumor budding and adjacent tissue at the invasive front can be potential novel predictors of DNM in early tongue cancer. Methods In total, 337 patients with early-stage tongue squamous cell carcinoma were retrospectively reviewed. The patient characteristics and histopathological factors were evaluated for association with DNM. DNM rates were calculated; items which were significant in the univariate analysis were used as explanatory variables, and independent factors for DNM were identified by the multivariate analysis. Results The univariate analysis identified T classification, depth of invasion, tumor budding, vascular invasion, and adjacent tissue at the invasive front as significant predictors of DNM; the multivariate analysis using these factors revealed all the above variables except vascular invasion, which are independent predictors of DNM. Conclusion In addition to conventional predictors, high grade tumor budding and adjacent tissue at the invasive front can serve as useful predictors of DNM in early tongue cancer.

de Melo N.B., Bernardino Í.D., de Melo D.P., Gomes D.Q., Bento P.M.

The goal of the present study was to investigate health-related quality of life (HRQoL) and oral health-related quality of life (OHRQoL) in patients under treatment for head and neck cancer and to identify the associated factors.A cross-sectional study was conducted with 102 patients undergoing treatment for head and neck cancer at 2 medical centers. Participants answered a sociodemographic questionnaire and the Brazilian versions of the Medical Outcomes Study 36 (SF-36) and the Oral Health Impact Profile (OHIP-14) questionnaires to assess HRQoL and OHRQoL, respectively. Clinical aspects, cancer staging, and treatment approach were also investigated. Descriptive and multivariate analyses were performed by using decision tree analysis with the Chi-square Automatic Interaction Detector (CHAID) algorithm.The decision tree revealed that reduced quality of life is associated with the clinical staging (adjusted P value = .035), patient's gender (adjusted P value = .028), and treatment approach (adjusted P value = .032). Female patients who are diagnosed with advanced head and neck cancer and undergo radiotherapy or chemotherapy are more likely to exhibit lower rates of quality of life.The results suggested that sociodemographic characteristics, clinical staging, and treatment approach can exert a significant influence on the quality of life of patients with head and neck cancer.

Yang X., Tian X., Wu K., Liu W., Li S., Zhang Z., Zhang C.

Although perineural invasion (PNI) has been recognized as a poor prognostic factor for oral cancer, few studies have focused on tongue squamous cell carcinoma (TSCC). Using a prospective randomized trial, this study investigated the role of PNI in the regional control and survival of the patients with cT1-2N0 TSCC, and clarified the benefit of neck management based on PNI status.PNI status was reviewed under H&E staining in tumors of 221 patients with cT1-2N0 TSCC, who were randomly assigned into elective neck dissection (END) group (n = 111) and observation group (n = 110). Oncologic and survival outcomes were analyzed by multivariate regression and Kaplan-Meier analyses.PNI was identified in 34 patients and multivariate analyses revealed that PNI remained an independent predictor for cervical lymph node metastasis (CLNM), local relapse, neck relapse and disease-specific survival (DSS) after controlling for T stage and pathologic differentiation. END could not improve the benefit for patients. Stratified analysis revealed that END also could not improve neck control or DSS among patients with PNI.This study demonstrated that PNI was an invaluable pathological parameter to independently predict cervical metastasis, local relapse, neck relapse and poor survival outcomes, but END could not improve benefits compared to observation for the PNI-positive patients.

Berdugo J., Thompson L.D., Purgina B., Sturgis C.D., Tuluc M., Seethala R., Chiosea S.I.

The 8th edition of American Joint Committee on Cancer (AJCC 8th) staging manual incorporated depth of invasion (DOI) into pT stage of oral cavity cancer. The aim of this study was to characterize several histological findings that may complicate measurement of DOI in early conventional squamous cell carcinomas (SCC) of the oral tongue: (1) lack of or minimal residual carcinoma following biopsy; (2) positive deep margin; (3) extratumoral perineural invasion (PNI); and (4) lymphatic or vascular invasion. Conventional SCC of the oral tongue (n = 407) with the largest dimension of ≤ 4 cm and with a negative elective cervical lymph node dissection (pN0) were reviewed. A clear plastic ruler was used to measure DOI by dropping a “plumb line” to the deepest point of the invasive tumor from the level of the basement membrane of the normal mucosa closest to the invasive tumor. Examples of identifying  reference point on the mucosal surface of oral tongue from which to measure the DOI are illustrated. In the experience of one contributing institution, the residual carcinoma was absent in 14.2% of glossectomies (34/239), while in 4.8% of cases (10/205) there was only minimal residual carcinoma. In 11.5% (21/183) of pT2 cases the deep margin was positive and thus DOI and pT may be underestimated. Of all cases with PNI, extratumoral PNI was identified in 23.1% (31/134) of cases, but represented the deepest point of invasion in only two cases. In one case, lymphatic invasion represented the deepest point of invasion and could have led to upstaging from pT1 to pT2. In conclusion, DOI measurement for SCC of the oral tongue may require re-examination of the diagnostic biopsy in up to 20% of cases due to the absence or only minimal residual carcinoma in glossectomy specimens. In 11.5% of apparently pT2 cases, DOI may be underestimated due to the positive deep margin. Rarely, extratumoral PNI or lymphatic invasion may be the deepest point of invasion. Overall, two issues (absent or minimal residual disease and positive deep margin) may confound DOI measurement in early SCCs of oral tongue.

Lynch C.M., Abdollahi B., Fuqua J.D., de Carlo A.R., Bartholomai J.A., Balgemann R.N., van Berkel V.H., Frieboes H.B.

Outcomes for cancer patients have been previously estimated by applying various machine learning techniques to large datasets such as the Surveillance, Epidemiology, and End Results (SEER) program database. In particular for lung cancer, it is not well understood which types of techniques would yield more predictive information, and which data attributes should be used in order to determine this information. In this study, a number of supervised learning techniques is applied to the SEER database to classify lung cancer patients in terms of survival, including linear regression, Decision Trees, Gradient Boosting Machines (GBM), Support Vector Machines (SVM), and a custom ensemble. Key data attributes in applying these methods include tumor grade, tumor size, gender, age, stage, and number of primaries, with the goal to enable comparison of predictive power between the various methods The prediction is treated like a continuous target, rather than a classification into categories, as a first step towards improving survival prediction. The results show that the predicted values agree with actual values for low to moderate survival times, which constitute the majority of the data. The best performing technique was the custom ensemble with a Root Mean Square Error (RMSE) value of 15.05. The most influential model within the custom ensemble was GBM, while Decision Trees may be inapplicable as it had too few discrete outputs. The results further show that among the five individual models generated, the most accurate was GBM with an RMSE value of 15.32. Although SVM underperformed with an RMSE value of 15.82, statistical analysis singles the SVM as the only model that generated a distinctive output. The results of the models are consistent with a classical Cox proportional hazards model used as a reference technique. We conclude that application of these supervised learning techniques to lung cancer data in the SEER database may be of use to estimate patient survival time with the ultimate goal to inform patient care decisions, and that the performance of these techniques with this particular dataset may be on par with that of classical methods.

Safi A., Kauke M., Grandoch A., Nickenig H., Zöller J.E., Kreppel M.

Recurrence is one of the main reasons for poor prognosis of OSCC. The mortality rate is approximately 90% and the 5-year overall survival rate decreases from 90% to 30% when recurrence is diagnosed. Identification of clinicopathological risk factors predicting recurrence may be helpful for patient individualized management and improvement of therapy. Therefore we investigated in our study the incidence of locoregional recurrences and their association with clinicopathological factors to identify possible significant risk factors.Our retrospective study consisted of 517 patients, who were diagnosed and treated between 2003-2013 at the Department for Oral and Maxillofacial Plastic Surgery, University of Cologne. Inclusion criteria were patients with treatment naive oral squamous cell carcinoma and primarily curative intended surgery with negative resection margins. Contingency tables and χ2-test were performed to analyse associations between clinicopathological features and recurrence. Multivariate analysis was performed using binary logistic regression analysis.We found out a significant correlation in univariate analysis between locoregional recurrence and number of resected cervical lymph nodes (p=0.013), number of positive cervical lymph nodes (p=0.041), postoperative radiatio (p=0.018), extracapsular spread (p=0.028) as well as grading (p=0.016). In multivariate analysis only grading was shown as independent risk factor for recurrence.Histological grading has been demonstrated as an independent risk factor for locoregional recurrence in the multivariate analysis. Furthermore, univariate analysis indicated the number of resected and positive lymph nodes, postoperative radiatio and extracapsular spread as significant risk factors. Taking these results into account, the mentioned parameters, especially histological grading, need to be considered for an individualized therapy management of patients with OSCC.

Zhang B., He X., Ouyang F., Gu D., Dong Y., Zhang L., Mo X., Huang W., Tian J., Zhang S.

We aimed to identify optimal machine-learning methods for radiomics-based prediction of local failure and distant failure in advanced nasopharyngeal carcinoma (NPC). We enrolled 110 patients with advanced NPC. A total of 970 radiomic features were extracted from MRI images for each patient. Six feature selection methods and nine classification methods were evaluated in terms of their performance. We applied the 10-fold cross-validation as the criterion for feature selection and classification. We repeated each combination for 50 times to obtain the mean area under the curve (AUC) and test error. We observed that the combination methods Random Forest (RF) + RF (AUC, 0.8464 ± 0.0069; test error, 0.3135 ± 0.0088) had the highest prognostic performance, followed by RF + Adaptive Boosting (AdaBoost) (AUC, 0.8204 ± 0.0095; test error, 0.3384 ± 0.0097), and Sure Independence Screening (SIS) + Linear Support Vector Machines (LSVM) (AUC, 0.7883 ± 0.0096; test error, 0.3985 ± 0.0100). Our radiomics study identified optimal machine-learning methods for the radiomics-based prediction of local failure and distant failure in advanced NPC, which could enhance the applications of radiomics in precision oncology and clinical practice.

Mroueh R., Haapaniemi A., Grénman R., Laranne J., Pukkila M., Almangush A., Salo T., Mäkitie A.

Incidence rates for oral tongue squamous cell carcinoma (SCC) are steadily rising worldwide.All patients diagnosed with primary oral tongue SCC at the 5 university hospitals in Finland from 2005 to 2009 were studied. The mean follow-up time was 43 months (median, 54 months; range, 0-111 months).Three hundred sixty patients with primary oral tongue SCC were identified. Treatment with curative intent was provided for 328 patients (91%). The 5-year disease-specific survival (DSS) rates were as follows: stage I 87%; stage II 73%; stage III 69%; and stage IV 51%. The 5-year recurrence-free survival in general has improved from 47% in our previous published series (1995-1999) to 65% in the current series (p < .001).The outcome of oral tongue SCC has significantly improved in Finland. However, the relatively high number of disease recurrences in patients with stage I and II disease, when compared with patients with stage III and IV disease, calls for an investigation of new treatment approaches. © 2017 Wiley Periodicals, Inc. Head Neck 39: 1306-1312, 2017.

Dwivedi A.K.

Diabetes as a chronic disease is becoming a foremost community health concern worldwide. In developing countries, the diabetic patients are increasing rapidly due to lack of sentience and bad eating habits. So, there is a need of a framework that can effectively diagnose thousands of patients using clinical specifics. This work uses six computational intelligence techniques for diabetes mellitus prediction namely classification tree, support vector machine, logistic regression, naïve Bayes, and artificial neural network. The performance of these techniques was evaluated on eight different classification performance measurements. Moreover, these techniques were appraised on a receiver operative characteristic curve. Classification accuracy of 77 and 78% was achieved by artificial neural network and logistic regression, respectively, with F 1 measure of 0.83 and 0.84.

Tota J.E., Anderson W.F., Coffey C., Califano J., Cozen W., Ferris R.L., St. John M., Cohen E.E., Chaturvedi A.K.

Despite significant reductions in tobacco use in the US, oral tongue cancer incidence has reportedly increased in recent years, particularly in young white women. We conducted age-period-cohort analyses to identify birth cohorts that have experienced increased oral tongue cancer incidence, and compared these with trends for oropharyngeal cancer, a cancer caused by human papillomavirus (HPV) that has also recently increased.We utilized cancer incidence data (1973-2012) from 18 registries maintained by the NCI SEER Program. Incidence trends were evaluated using log-linear joinpoint regression and age-period-cohort modeling was utilized to simultaneously evaluate effects of age, calendar year, and birth year on incidence trends.Incidence of oral tongue cancer increased significantly among white women during 1973-2012 (0.6% annual increase, p

Arora A., Husain N., Bansal A., Neyaz A., Jaiswal R., Jain K., Chaturvedi A., Anand N., Malhotra K., Shukla S.

The aim of this study was to evaluate the histopathologic parameters that predict lymph node metastasis in patients with oral squamous cell carcinoma (OSCC) and to design a new assessment score on the basis of these parameters that could ultimately allow for changes in treatment decisions or aid clinicians in deciding whether there is a need for close follow-up or to perform early lymph node dissection. Histopathologic parameters of 336 cases of OSCC with stage cT1/T2 N0M0 disease were analyzed. The location of the tumor and the type of surgery used for the management of the tumor were recorded for all patients. The parameters, including T stage, grading of tumor, tumor budding, tumor thickness, depth of invasion, shape of tumor nest, lymphoid response at tumor-host interface and pattern of invasion, eosinophilic reaction, foreign-body giant cell reaction, lymphovascular invasion, and perineural invasion, were examined. Ninety-two patients had metastasis in lymph nodes. On univariate and multivariate analysis, independent variables for predicting lymph node metastasis in descending order were depth of invasion (P=0.003), pattern of invasion (P=0.007), perineural invasion (P=0.014), grade (P=0.028), lymphovascular invasion (P=0.038), lymphoid response (P=0.037), and tumor budding (P=0.039). We designed a scoring system on the basis of these statistical results and tested it. Cases with scores ranging from 7 to 11, 12 to 16, and ≥17 points showed LN metastasis in 6.4%, 22.8%, and 77.1% of cases, respectively. The difference between these 3 groups in relation to nodal metastasis was very significant (P

Lydiatt W.M., Patel S.G., O'Sullivan B., Brandwein M.S., Ridge J.A., Migliacci J.C., Loomis A.M., Shah J.P.

Answer questions and earn CME/CNE The recently released eighth edition of the American Joint Committee on Cancer (AJCC) Staging Manual, Head and Neck Section, introduces significant modifications from the prior seventh edition. This article details several of the most significant modifications, and the rationale for the revisions, to alert the reader to evolution of the field. The most significant update creates a separate staging algorithm for high-risk human papillomavirus-associated cancer of the oropharynx, distinguishing it from oropharyngeal cancer with other causes. Other modifications include: the reorganizing of skin cancer (other than melanoma and Merkel cell carcinoma) from a general chapter for the entire body to a head and neck-specific cutaneous malignancies chapter; division of cancer of the pharynx into 3 separate chapters; changes to the tumor (T) categories for oral cavity, skin, and nasopharynx; and the addition of extranodal cancer extension to lymph node category (N) in all but the viral-related cancers and mucosal melanoma. The Head and Neck Task Force worked with colleagues around the world to derive a staging system that reflects ongoing changes in head and neck oncology; it remains user friendly and consistent with the traditional tumor, lymph node, metastasis (TNM) staging paradigm. CA Cancer J Clin 2017;67:122-137. © 2017 American Cancer Society.

Ng J.H., Iyer N.G., Tan M., Edgren G.

There are reports about the changing epidemiology of tongue squamous cell carcinoma (SCC), with recent reports indicating an increasing incidence in young women.Data on incident cases of tongue SCC were collected from cancer registries worldwide.Data from a total of 22 cancer registries and 89,212 incident cases of tongue SCC worldwide were included. Most areas experienced an incidence increase ranging from 0.4% to 3.3% per year. There was a significant difference in the incidence increase between sexes in 11 of the 22 registries. In 14 of the 22 registries studied, the increase in incidence of tongue SCC was higher in the group of subjects

Gupta A., Bajaj S., Nema P., Purohit A., Kashaw V., Soni V., Kashaw S.K.

Lunitsyna Y.V., Shevyakina A.O., Tokmakova S.I., Bondarenko O.V.

Relevance. Squamous cell carcinoma of the head and neck, including oral and oropharyngeal cancer, ranks as the seventh most common cancer worldwide, contributing to over 660,000 new cases and 325,000 deaths annually. Understanding the interplay of adverse factors, their association with disease development, and the creation of mathematical risk prediction models can play a crucial role in enhancing screening efforts and advancing primary prevention of malignant neoplasms.Objective. This study aimed to identify significant risk factors for oral mucosa cancer in the Altai Krai population and to develop a mathematical model for disease risk assessment.Material and methods. The study included 184 patients diagnosed with oral mucosa cancer, along with a control group of 416 healthy volunteers with no history or current diagnosis of oncological diseases. A total of 39 potential risk factors were analyzed across all participants. Statistical analyses were conducted to identify region-specific risk factors for Altai Krai. Binary logistic regression and ROC analysis were used to construct the risk prediction model.Results. Comparative analysis between the patient group and the control group revealed differences in 19 of the 39 evaluated factors. However, the final risk prediction model identified five key factors significantly influencing disease development. Advanced age, smoking status, the number of cigarettes smoked, and alcohol consumption were found to substantially increase the risk of oral mucosa cancer, while engagement in intellectual work-related activities was associated with a reduced risk. The resulting formula demonstrated high predictive accuracy, with an Area Under the ROC Curve (AUC) of 0.91, a standard error of 0.024, and a 95% confidence interval of 0.856–0.955 (z-statistic: 17.50; significance level: P (area = 0.5) < 0.001). Both the sensitivity and specificity of the model were high.Conclusion. The developed risk assessment model shows great promise for helping screen and diagnose oral cancer in the Altai Krai population. This tool could give dentists and healthcare providers a simple, practical way to identify individuals at risk early by using well-established risk factors.

Alabi R.O., Elmusrati M., Leivo I., Almangush A., Mäkitie A.A.

Alam M.I., Khan H., Alam M.Z., Siddiqui S.T., Haider A.S., Khan M.R.

When it comes to medical studies and the life sciences, Machine Learning already made a significant impact. The metabolic disease known as diabetes is characterized by continuously high blood sugar levels that do not respond normally to insulin. Early diagnosis of diabetes helps to maintain a healthy lifestyle. The article’s content has centered on analyzing PIMA dataset-based diabetes patients and developing a machine learning-based detection model with minimal dependencies. Machine learning (ML) algorithms will be an effective strategy because they can be trained and tested using large amounts of data and can further improve themselves by making predictions. Several algorithms, including Gradient Boosting, Decision Tree, Random Forest, Support Vector Machine (SVM), K-Nearest Neighbours (KNN), and Naive Bayes, are trained using our collected dataset in this article. Random Forest’s prediction results are shown to be the most accurate after being compared to those of the other algorithms.

Du W., Jia M., Li J., Gao M., Zhang W., Yu Y., Wang H., Peng X.

Although rare overall, salivary gland carcinomas (SGCs) are among the most common oral and maxillofacial malignancies. The aim of this study was to develop a machine learning-based model to predict the survival of patients with SGC. Patients in whom SGC was confirmed by histological testing and who underwent primary extirpation at the authors' institution between 1963 and 2014 were identified. Demographic and clinicopathological data with complete follow-up information were collected for analysis. Feature selection methods were used to determine the correlation between prognosis-related factors and survival in the collected patient data. The collected clinicopathological data and multiple machine learning algorithms were used to develop a survival prediction model. Three machine learning algorithms were applied to construct the prediction models. The area under the receiver operating characteristic curve (AUC) and accuracy were used to measure model performance. The best classification performance was achieved with a LightGBM algorithm (AUC = 0.83, accuracy = 0.91). This model enabled prognostic prediction of patient survival. The model may be useful in developing personalized diagnostic and treatment strategies and formulating individualized follow-up plans, as well as assisting in the communication between doctors and patients, facilitating a better understanding of and compliance with treatment.

Omobolaji Alabi R., Mäkitie R.E.

Jeter R., Greenfield R., Housley S.N., Belykh I.

Background Stroke therapy is essential to reduce impairments and improve motor movements by engaging autogenous neuroplasticity. Traditionally, stroke rehabilitation occurs in inpatient and outpatient rehabilitation facilities. However, recent literature increasingly explores moving the recovery process into the home and integrating technology-based interventions. This study advances this goal by promoting in-home, autonomous recovery for patients who experienced a stroke through robotics-assisted rehabilitation and classifying stroke residual severity using machine learning methods. Objective Our main objective is to use kinematics data collected during in-home, self-guided therapy sessions to develop supervised machine learning methods, to address a clinician’s autonomous classification of stroke residual severity–labeled data toward improving in-home, robotics-assisted stroke rehabilitation. Methods In total, 33 patients who experienced a stroke participated in in-home therapy sessions using Motus Nova robotics rehabilitation technology to capture upper and lower body motion. During each therapy session, the Motus Hand and Motus Foot devices collected movement data, assistance data, and activity-specific data. We then synthesized, processed, and summarized these data. Next, the therapy session data were paired with clinician-informed, discrete stroke residual severity labels: “no range of motion (ROM),” “low ROM,” and “high ROM.” Afterward, an 80%:20% split was performed to divide the dataset into a training set and a holdout test set. We used 4 machine learning algorithms to classify stroke residual severity: light gradient boosting (LGB), extra trees classifier, deep feed-forward neural network, and classical logistic regression. We selected models based on 10-fold cross-validation and measured their performance on a holdout test dataset using F1-score to identify which model maximizes stroke residual severity classification accuracy. Results We demonstrated that the LGB method provides the most reliable autonomous detection of stroke severity. The trained model is a consensus model that consists of 139 decision trees with up to 115 leaves each. This LGB model boasts a 96.70% F1-score compared to logistic regression (55.82%), extra trees classifier (94.81%), and deep feed-forward neural network (70.11%). Conclusions We showed how objectively measured rehabilitation training paired with machine learning methods can be used to identify the residual stroke severity class, with efforts to enhance in-home self-guided, individualized stroke rehabilitation. The model we trained relies only on session summary statistics, meaning it can potentially be integrated into similar settings for real-time classification, such as outpatient rehabilitation facilities.

Alapati R., Renslo B., Wagoner S.F., Karadaghy O., Serpedin A., Kim Y.E., Feucht M., Wang N., Ramesh U., Bon Nieves A., Lawrence A., Virgen C., Sawaf T., Rameau A., Bur A.M.

ObjectiveThis study aimed to assess reporting quality of machine learning (ML) algorithms in the head and neck oncology literature using the TRIPOD‐AI criteria.Data SourcesA comprehensive search was conducted using PubMed, Scopus, Embase, and Cochrane Database of Systematic Reviews, incorporating search terms related to “artificial intelligence,” “machine learning,” “deep learning,” “neural network,” and various head and neck neoplasms.Review MethodsTwo independent reviewers analyzed each published study for adherence to the 65‐point TRIPOD‐AI criteria. Items were classified as “Yes,” “No,” or “NA” for each publication. The proportion of studies satisfying each TRIPOD‐AI criterion was calculated. Additionally, the evidence level for each study was evaluated independently by two reviewers using the Oxford Centre for Evidence‐Based Medicine (OCEBM) Levels of Evidence. Discrepancies were reconciled through discussion until consensus was reached.ResultsThe study highlights the need for improvements in ML algorithm reporting in head and neck oncology. This includes more comprehensive descriptions of datasets, standardization of model performance reporting, and increased sharing of ML models, data, and code with the research community. Adoption of TRIPOD‐AI is necessary for achieving standardized ML research reporting in head and neck oncology.ConclusionCurrent reporting of ML algorithms hinders clinical application, reproducibility, and understanding of the data used for model training. To overcome these limitations and improve patient and clinician trust, ML developers should provide open access to models, code, and source data, fostering iterative progress through community critique, thus enhancing model accuracy and mitigating biases.Level of EvidenceNA Laryngoscope, 2024

AlRashdi S., AlHassani A., Haile F., AlNuaimi R., Labben T., Ertek G.

In this study, the Electric Vehicle (EV) purchase decisions of European consumers are predicted using supervised machine learning (ML), specifically classification. Following the replacement (imputing) of missing data values through predicted values and continuizing of all predictor features, the predictor features are ranked according to the Information Gain Ratio and the Gini coefficient. The results suggest that suiting daily driving needs (Q17), belief that society must reward electric cars instead of petrol and diesel cars (Q14), and opinion change regarding electric cars during the past year (Q21) ranked the highest with respect to the Gini coefficient metric. The same predictor features rank the highest with respect to the Information Gain Ratio metric, yet in a different rank (Q17, Q21, and Q14). For predictive analytics, a multitude of classification algorithms are applied to predict the decision of EV purchase, and the performance of the applied algorithms is compared. The results suggest that gradient boosting performed best in predicting EV adoption decisions, followed by the logistic regression and random forest algorithms.

Dong F., Yan J., Zhang X., Zhang Y., Liu D., Pan X., Xue L., Liu Y.

Application of deep learning (DL) and machine learning (ML) is rapidly increasing in the medical field. DL is gaining significance for medical image analysis, particularly, in oral and maxillofacial surgeries. Owing to the ability to accurately identify and categorize both diseased and normal soft- and hard-tissue structures, DL has high application potential in the diagnosis and treatment of tumors and in orthognathic surgeries. Moreover, DL and ML can be used to develop prediction models that can aid surgeons to assess prognosis by analyzing the patient's medical history, imaging data, and surgical records, develop more effective treatment strategies, select appropriate surgical modalities, and evaluate the risk of postoperative complications. Such prediction models can play a crucial role in the selection of treatment strategies for oral and maxillofacial surgeries. Their practical application can improve the utilization of medical staff, increase the treatment accuracy and efficiency, reduce surgical risks, and provide an enhanced treatment experience to patients. However, DL and ML face limitations, such as data drift, unstable model results, and vulnerable social trust. With the advancement of social concepts and technologies, the use of these models in oral and maxillofacial surgery is anticipated to become more comprehensive and extensive.

Omobolaji Alabi R., Elmusrati M., Leivo I., Almangush A., Mäkitie A.A.

Radiomics is a rapidly growing field used to leverage medical radiological images by extracting quantitative features. These are supposed to characterize a patient's phenotype, and when combined with artificial intelligence techniques, to improve the accuracy of diagnostic models and clinical outcome prediction. This review aims at examining the application areas of artificial intelligence-based radiomics (AI-based radiomics) for the management of head and neck cancer (HNC). It further explores the workflow of AI-based radiomics for personalized and precision oncology in HNC. Finally, it examines the current challenges of AI-based radiomics in daily clinical oncology and offers possible solutions to these challenges. Comprehensive electronic databases (PubMed, Medline via Ovid, Scopus, Web of Science, CINAHL, and Cochrane Library) were searched following the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines. The quality of included studies and their risk of biases were evaluated using the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) and Prediction Model Risk of Bias Assessment Tool (PROBAST). Out of the 659 search hits retrieved, 45 fulfilled the inclusion criteria. Our review revealed that the application of AI-based radiomics model as an ancillary tool for improved decision-making in HNC management includes radiomics-based cancer diagnosis and radiomics-based cancer prognosis. The radiomics-based cancer diagnosis includes tumor staging, tumor grading, and classification of malignant and benign tumors. Similarly, radiomics-based cancer prognosis includes prediction for treatment response, recurrence, metastasis, and survival. In addition, the challenges in the implementation of these models for clinical evaluations include data imbalance, feature engineering (extraction and selection), model generalizability, multi-modal fusion, and model interpretability. Considering the highly subjective and interobserver variability that is peculiar to the interpretation of medical images by expert clinicians, AI-based radiomics seeks to offer potentially useful quantitative information, which is not visible to the human eye or unintentionally often remain ignored during clinical imaging practice. By enabling the extraction of this type of information, AI-based radiomics has the potential to revolutionize HNC oncology, providing a platform for more personalized, higher quality, and cost-effective care for HNC patients.

Moharrami M., Azimian Zavareh P., Watson E., Singhal S., Johnson A.E., Hosni A., Quinonez C., Glogauer M.

Background This systematic review aimed to evaluate the performance of machine learning (ML) models in predicting post-treatment survival and disease progression outcomes, including recurrence and metastasis, in head and neck cancer (HNC) using clinicopathological structured data. Methods A systematic search was conducted across the Medline, Scopus, Embase, Web of Science, and Google Scholar databases. The methodological characteristics and performance metrics of studies that developed and validated ML models were assessed. The risk of bias was evaluated using the Prediction model Risk Of Bias ASsessment Tool (PROBAST). Results Out of 5,560 unique records, 34 articles were included. For survival outcome, the ML model outperformed the Cox proportional hazards model in time-to-event analyses for HNC, with a concordance index of 0.70–0.79 vs. 0.66–0.76, and for all sub-sites including oral cavity (0.73–0.89 vs. 0.69–0.77) and larynx (0.71–0.85 vs. 0.57–0.74). In binary classification analysis, the area under the receiver operating characteristics (AUROC) of ML models ranged from 0.75–0.97, with an F1-score of 0.65–0.89 for HNC; AUROC of 0.61–0.91 and F1-score of 0.58–0.86 for the oral cavity; and AUROC of 0.76–0.97 and F1-score of 0.63–0.92 for the larynx. Disease-specific survival outcomes showed higher performance than overall survival outcomes, but the performance of ML models did not differ between three- and five-year follow-up durations. For disease progression outcomes, no time-to-event metrics were reported for ML models. For binary classification of the oral cavity, the only evaluated subsite, the AUROC ranged from 0.67 to 0.97, with F1-scores between 0.53 and 0.89. Conclusions ML models have demonstrated considerable potential in predicting post-treatment survival and disease progression, consistently outperforming traditional linear models and their derived nomograms. Future research should incorporate more comprehensive treatment features, emphasize disease progression outcomes, and establish model generalizability through external validations and the use of multicenter datasets.

Wahab A., Bello I.O., Alabi R.O., Mascitti M., Troiano G., Mauramo M., Coletta R.D., Salo T., Almangush A.

AbstractBackgroundOral tongue squamous cell carcinoma (OTSCC) often presents with aggressive clinical behaviour that may require multimodality treatment based on reliable prognostication. We aimed to evaluate the prognostic ability of five online web‐based tools to predict the clinical behaviour of OTSCC resection and biopsy samples.MethodsA total of 135 OTSCC resection cases and 33 OTSCC biopsies were included to predict recurrence and survival. Area under the receiver operating characteristic curves (AUC), χ2 tests, and calibration plots constructed to estimate the prognostic power of each tool.ResultsThe tool entitled ‘Prediction of risk of Locoregional Recurrences in Early OTSCC’ presented an accuracy of 82%. The tool, ‘Head & Neck Cancer Outcome Calculator’ for 10‐year cancer‐related mortality had an accuracy 77% and AUC 0.858. The other tool entitled ‘Cancer Survival Rates’ for 5‐year mortality showed an accuracy of 74% and AUC of 0.723. For biopsy samples, ‘Cancer Survival Prediction Calculators’ predicted the recurrence free survival with an accuracy of 70%.ConclusionsWeb‐based tools can aid in clinical decision making of OTSCC. Three of five online web‐based tools could predict recurrence risk and cancer‐related mortality in resected OTSCC and one tool could help in clinical decision making for biopsy samples.

Ma K., Qiao L., Lin G., Xing G.

Uppal S., Kumar Shrivastav P., Khan A., Sharma A., Kumar Shrivastava A.

Oral Potentially Malignant Disorders (OPMDs) refer to a heterogenous group of clinical presentations with heightened rate of malignant transformation. Identification of risk levels in OPMDs is crucial to determine the need for active intervention in high-risk patients and routine follow-up in low-risk ones. Machine learning models has shown tremendous potential in several areas of dentistry that strongly suggest its application to estimate rate of malignant transformation of precancerous lesions. A comprehensive literature search was performed on Pubmed/MEDLINE, Web of Science, Scopus, Embase, Cochrane Library database to identify articles including machine learning models and algorithms to predict malignant transformation in OPMDs. Relevant bibliographic data, study characteristics, and outcomes were extracted for eligible studies. Quality of the included studies was assessed through the IJMEDI checklist. Fifteen articles were found suitable for the review as per the PECOS criteria. Amongst all studies, highest sensitivity (100%) was recorded for U-net architecture, Peaks Random forest model, and Partial least squares discriminant analysis (PLSDA). Highest specificity (100%) was noted for PLSDA. Range of overall accuracy in risk prediction was between 95.4% and 74%. Machine learning proved to be a viable tool in risk prediction, demonstrating heightened sensitivity, automation, and improved accuracy for predicting transformation of OPMDs. It presents an effective approach for incorporating multiple variables to monitor the progression of OPMDs and predict their malignant potential. However, its sensitivity to dataset characteristics necessitates the optimization of input parameters to maximize the efficiency of the classifiers.