The Cox proportional hazards model commonly used to evaluate prognostic variables in survival of cancer patients may be too simplistic to properly predict a cancer patient’s outcome since it assumes that the outcome is a linear combination of covariates. In this retrospective study including 255 patients suitable for analysis who underwent surgical treatment in our department from 2000 to 2017, we applied a deep learning-based survival prediction method in oral squamous cell carcinoma (SCC) patients and validated its performance. Survival prediction using DeepSurv, a deep learning based-survival prediction algorithm, was compared with random survival forest (RSF) and the Cox proportional hazard model (CPH). DeepSurv showed the best performance among the three models, the c-index of the training and testing sets reaching 0.810 and 0.781, respectively, followed by RSF (0.770/0.764), and CPH (0.756/0.694). The performance of DeepSurv steadily improved with added features. Thus, deep learning-based survival prediction may improve prediction accuracy and guide clinicians both in choosing treatment options for better survival and in avoiding unnecessary treatments.

Patients with odontogenic cysts and tumors may have to undergo serious surgery unless the lesion is properly detected at the early stage. The purpose of this study is to evaluate the diagnostic performance of the real-time object detecting deep convolutional neural network You Only Look Once (YOLO) v2—a deep learning algorithm that can both detect and classify an object at the same time—on panoramic radiographs. In this study, 1602 lesions on panoramic radiographs taken from 2010 to 2019 at Yonsei University Dental Hospital were selected as a database. Images were classified and labeled into four categories: dentigerous cysts, odontogenic keratocyst, ameloblastoma, and no lesion. Comparative analysis among three groups (YOLO, oral and maxillofacial surgeons, and general practitioners) was done in terms of precision, recall, accuracy, and F1 score. While YOLO ranked highest among the three groups (precision = 0.707, recall = 0.680), the performance differences between the machine and clinicians were statistically insignificant. The results of this study indicate the usefulness of auto-detecting convolutional networks in certain pathology detection and thus morbidity prevention in the field of oral and maxillofacial surgery.

It has been recognized that systemic inflammatory markers (SIMs) are associated with patient survival in various types of cancer. This study aimed to determine the optimal cut-off values, and to evaluate the prognostic performance of SIMs for oral squamous cell carcinoma (OSCC) within the framework of the American Joint Committee of Cancer (AJCC) cancer staging manual, 8th edition. Records were collected for a total 291 patients who had had a peripheral blood test within 1 week prior to surgery and had undergone the surgical resection of OSCC in a single institution between 2005 and 2018. The cut-off values of SIMs were obtained, and the survival analyses for overall survival (OS) and disease-free survival (DFS) were performed. Multivariate analyses incorporating other clinicopathologic factors were performed to verify the independent risk factors for survival. The cut-off values of neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR) were 2.23, 135.14 for OS and 2.16, 131.07 for DFS, respectively, demonstrating a significant association for OS and DFS in OSCC. AJCC pathologic regional lymph node category (pN) (P < 0.001), perineural invasion (PNI) (P < 0.001) and NLR (P < 0.001) were independent predictors for OS. Meanwhile, for DFS, AJCC pN (P = 0.018) and NLR (P = 0.015) were shown to be independent predictors. Before the curative surgery, NLR and PLR could be auxiliary parameters for OS and DFS in OSCC. And based on the 8th edition of AJCC staging system, elevated NLR will be a potential indicator of the worse OS or DFS along with pN or PNI in OSCC.