Role of Artificial Intelligence in Diagnosis of Oral Cancer: A Systematic Review

Survival rates are considerably raised when oral cancer (OC) is identified early. The application of artificial intelligence (AI) technology to diagnostic medicine has attracted more attention recently. The purpose of this study was to evaluate the evidence that is currently available on the effectiveness of AI in diagnosing OC. A special focus has been put on AI's diagnostic precision along with its ability to detect OC in its early phases. We conducted a systematic search on four different databases (Scopus, Web of Science, PubMed and IEEE Xpolore) to search for relevant studies. We found 592 studies which were retrieved to Endnote X9 software and duplicates were removed. After assessing all the studies on inclusion and exclusion criteria, 16 studies, comprising 69,425 photos and 7245 patients, were found relevant and were included in this systematic review. All these studies were assessed for potential risk of bias using QUADAS 2 tool. The included studies' AI performance was evaluated using ten statistical techniques. Eleven studies employed deep learning, while six used supervised machine learning. Deep learning (DL) yielded findings ranging from 81 to 99.7% accuracy, 79-98.75% sensitivity, 82-100% specificity, and 79-99.5% area under the curve (AUC). The accuracy, sensitivity, and specificity ranged from 43.5 to 100%, 94-100%, and 93%, respectively, according to the results of supervised machine learning. The optimal AI technique for OC detection is not universally agreed upon. AI is a useful diagnostic technique that marks a significant advancement in the early diagnosis of OC. According to the data, supervised machine learning is less reliable than DL, which includes a deep convolutional neural network, in the early identification of OC.