Oral Squamous Cell Carcinoma (OSCC) is a common type of cancer of the oral epithelium. Despite their high impact on mortality, sufficient screening methods for early diagnosis of OSCC often lack accuracy and thus OSCCs are mostly diagnosed at a late stage. Early detection and accurate outline estimation of OSCCs would lead to a better curative outcome and a reduction in recurrence rates after surgical treatment. Confocal Laser Endomicroscopy (CLE) records sub-surface micro-anatomical images for in vivo cell structure analysis. Recent CLE studies showed great prospects for a reliable, real-time ultrastructural imaging of OSCC in situ. We present and evaluate a novel automatic approach for OSCC diagnosis using deep learning technologies on CLE images. The method is compared against textural feature-based machine learning approaches that represent the current state of the art. For this work, CLE image sequences (7894 images) from patients diagnosed with OSCC were obtained from 4 specific locations in the oral cavity, including the OSCC lesion. The present approach is found to outperform the state of the art in CLE image recognition with an area under the curve (AUC) of 0.96 and a mean accuracy of 88.3% (sensitivity 86.6%, specificity 90%).

BackgroundConfocal laser endomicroscopy (CLE) is an optical biopsy method allowing in vivo microscopic imaging at 1000-fold magnification. It was the aim to evaluate CLE in the human oral cavity for the differentiation of physiological/carcinomatous mucosa and to establish and validate, for the first time, a scoring system to facilitate CLE assessment.MethodsThe study consisted of 4 phases: (1) CLE-imaging (in vivo) was performed after the intravenous injection of fluorescein in patients with histologically confirmed carcinomatous oral mucosa; (2) CLE-experts (n = 3) verified the applicability of CLE in the oral cavity for the differentiation between physiological and cancerous tissue compared to the gold standard of histopathological assessment; (3) based on specific patterns of tissue changes, CLE-experts (n = 3) developed a classification and scoring system (DOC-Score) to simplify the diagnosis of oral squamous cell carcinomas; (4) validation of the newly developed DOC-Score by non-CLE-experts (n = 3); final statistical evaluation of their classification performance (comparison to the results of CLE-experts and the histopathological analyses).ResultsExperts acquired and edited 45 sequences (260 s) of physiological and 50 sequences (518 s) of carcinomatous mucosa (total: 95 sequences/778 s). All sequences were evaluated independently by experts and non-experts (based on the newly proposed classification system). Sensitivity (0.953) and specificity (0.889) of the diagnoses by experts as well as sensitivity (0.973) and specificity (0.881) of the non-expert ratings correlated well with the results of the present gold standard of tissue histopathology. Experts had a positive predictive value (PPV) of 0.905 and a negative predictive value (NPV) of 0.945. Non-experts reached a PPV of 0.901 and a NPV of 0.967 with the help of the DOC-Score. Inter-rater reliability (Fleiss` kappa) was 0.73 for experts and 0.814 for non-experts. The intra-rater reliability (Cronbach’s alpha) of the experts was 0.989 and 0.884 for non-experts.ConclusionsCLE is a suitable and valid method for experts to diagnose oral cancer. Using the DOC-Score system, an accurate chair-side diagnosis of oral cancer is feasible with comparable results to the gold standard of histopathology—even in daily clinical practice for non-experienced raters.

BackgroundLocal anesthesia is an important skill and a prerequisite for most dental treatments. However, the step from theory to application on the patient is huge for the novice. Hence, a mannequin training model course was developed and implemented into the existing local anesthesia curriculum in undergraduate dental students. It was the aim of this study to evaluate the relation between training-model and real-life anesthesia performance and to measure whether a gain in skill on the model translates to the actual patient situation.MethodsThirty-six third-year students (14 males, 22 females, age 24 years±2.98) attended the four-day course comprising each 4 h of lectures and practical training. The student cohort gave subjective ratings about the didactical components of the course after attendance by using the TRIL questionnaire (TRIL-mod; University of Trier). At the end of the course the performance of each student in administering an inferior alveolar nerve (IAN) block on the training model as well as on a fellow dental student was investigated using a standardized checklist. To evaluate the successful performance, the in vivo IAN-block was assessed using subjective patient-feeling, the sharp-blunt test and an objective pain- and thermal sensitivity tester (PATH).ResultsThe course was rated with an average score of 5.25 ± 0.44 (range 1–6; 6 = best). On the training model, 69.4% of the students successfully performed an IAN-block. The in vivo assessment, objectified by the PATH test, showed a successful anesthesia in 36.9% of the cases. The assessment of local anesthesia by using the sharp blunt test and the subjective patient feeling significantly correlated with these findings (k = 0.453–0.751, p < 0.05). The model performance did not correlate with the performance on the patient (k = 0.137, p = 0.198).ConclusionsAlthough subjective ratings of the course were high, the anesthesia success rate on mannequin models did not imply an equal performance on the in vivo setting. As local anesthesia training models are a valuable didactic complement, the focus of the training should be on to the actual real life situation. Chair side feedback should be offered to the students using one of the presented evaluation methods.