Comments to the authors
Thank you for asking to review the manuscript entitled “Application of artificial intelligence and machine learning for prediction of oral cancer risk.” The authors have built an artificial intelligence (AI) tool based on artificial neural network (ANN) model to predict the risk of oral cancer development. The input dataset included 73 samples which were randomly split into training (54) and testing (19) groups to training and validating the algorithm. According to the authors, the model was approximately 86% sensitive, 60% specific, and 79% accurate. They have concluded that this ANN model could help in the clinical decision making within oral cancer context.
This is an interesting work, but some changes, including English grammatical review, are needed before the acceptance of the paper.
Abstract
1. Page 3; lines 6-10 "The aim of this study was to develop an artificial neural network capable of predicting oral cancer risk based on data on risk factors, systematic medical condition and clinic-pathological features."
Please, correct the typo in red. (correction: systemic)
Introduction
1. Page 3; lines 46-48 “Oral cancer (OC) is one among the most common cancer affecting the mankind. OC mainly includes the subsites of lips, oral cavity, nasopharynx and pharynx.”
Although this information was extracted from the World Health Organization website, I do not agree with this statement. Despite the oral cavity and oropharynx are within the head and neck region, tumors from these sites are clinic-pathologic distinct. They present different risk factors and are clinically and molecularly distinct. Therefore, this statement does not seem appropriate.
2. Page 4; line 19 “Artificial intelligence (AI) in one of the major breakthrough in the field of science and technology.”
Please, correct the typo in red. (correction: is)
3. Page 4; lines 38-41 “The input layer after receiving the data, transmits it to a hidden layer, which is used for processing the data and later on provides the results to the outer layer”
It is important to describe that one or more hidden layers may constitute an ANN model. The term "output layer" is more commonly used than "outer layer".
Materials and methods
1. Patient data
1.1 Page 5, lines 11-12. “73 cases which fulfilled the eligibility criteria’s were included in the study.”
Please, inform what subtypes of oral cancer were included.
2. Development of ANNs based prediction model
2.1 The use of AI models in oral oncology is on its initial stage. The works published in this field needs to be as clear as possible both to familiarize clinicians and pathologists with the topic and to provide the technical reproducibility. Therefore, some points need to be addressed here. For instance: how many neurons were used in the hidden layer and why? Why was only one hidden layer used in the model?
3. Training/Validating of ANNs based prediction model
3.1 Page 5, line 41. “The model was trained for 10,000 iterations”
Please, correct the typo in red. (Correction: interactions)
3.2 The cross-validation method used to assess the accuracy of the test set should be more detailed here.
4. Testing the performance of ANNs based prediction model
4.1 Page 6, line 9. “FN is when the model incorrectly predicts the positive cases.”
Please, correct the typo in red. (Correction: FP)
Results
1. Page 6; lines 17-20 “Of these 73 cases, 20 (27%) were benign cases, and 53 (73%) were malignant cases. 37 were male and 36 were female, with a mean age of 54.45 years as shown in Table 2.”
This statement is very confusing for the reader. What were these lesions? In the materials and methods, along the "patient data" section the authors stated: "In the present study patient data was obtained after a thorough search for oral cancer cases registered between 2017 and 2020 at oral pathology service at Prince Mohammed Bin Nasser Hospital. A total of 138 confirmed cases with pathologic reports were retrieved." This sentence leads the reader to believe that all were OC cases. Therefore, what types of benign cases the authors mean? Did the authors include oral potentially malignant disorders? All the cases included in this work should be specified. Additionally, the details about the samples showed in Table 2 do not include understanding information to the work. I suggest the authors add a new table with complete details about each case included in the analysis (the diagnosis of lesion, sex, age, anatomic location, clinical presentation, symptoms, comorbidities and lifestyle habits).
Discussion
1. Page 6, line 55. “Integration of variables like age, gender, family history, high risk habits play a crucial role in predicting the development of cancer.21,21”
Please, correct the citation.
2. Page 7, lines 33-34 “The limitation of this study is with the sample size, a small sample size was included in the study which might have affected the results.”
This is a true limitation, but the sentence is vague. How the sample size affects the results? It is an excellent opportunity for the authors to discuss a little bit about overfitting, underfitting and methods to overcome these issues.
Conclusion
Page 7, lines 39-43. “The results demonstrate that the ANN model could perform well in estimating the probability of malignancy, and improve the positive predictive value (PPV) of the decision to perform biopsy.”
The statement highlighted in red is one of my major concerns with this work. The decision to perform biopsy was not discussed in any moment of the text and it was not even mentioned as one of the aims of the work. How did the authors reach this conclusion? In my view, this is an extrapolation, as the present results do not support this conclusion. This work could only be considered for publishing if the authors include more information about the cases and explain in detail how this ANN model can guide the clinical decision making regarding the biopsy procedures.

After the corrections, the text became clearer and revealed a valuable study. Thus, I agree with the acceptance of the manuscript without further comments.