The classification of oral squamous cell carcinoma (OSCC) by means of transfer learning

Patients that are diagnosed with oral cancer has more than an 83% survival chance if it is detected in its early stages. However, through conventional labour-intensive means, only 29% of cases are detected. It is worth mentioning that 90% of oral cancer is Oral Squamous Cell Carcinoma (OSCC) and is often caused by smoking and alcohol consumption. Computer-aided diagnostics could further increase the rate of detection of this form of oral cancer. The present study sought to employ a class of deep learning techniques known as transfer learning. The Inception V3 pre-trained convolutional neural network model is used to extract the features from texture-based images. Consequently, the malignant and benign nature of the cancer is identified from three different machine learning models, i.e., Support Vector Machine (SVM), k-Nearest Neighbors (kNN) and Random Forest (RF). It was shown from the study that an average of 91% classification accuracy was obtained from the test and validation dataset from the Inception V3-RF pipeline. The outcome of the present study could serve useful in an objective-based automatic diagnostic of OSCC and hence could possibly increase its detection