Search

Recent advancements in machine learning techniques for salivary gland tumors have shown promising results, favoring image analysis to differentiate between malignant and benign SGT. Computational vision holds potential in supporting pathologists by aiding in clinical decision-making processes, including biopsy referrals. In this context, preprocessing steps such as segmentation of clinical photographs can enhance the Convolutional Neural Network performance by reducing noise compared to frameworks that process images without segmentation. The aim of this investigation is to implement and evaluate the potential of a CNN segmentation model – UNet – to segment photographs of salivary gland tumors. A dataset consisting of 100.JPG images with corresponding.XML annotations was converted into.JPG masks for analysis. A subset of 20 images was randomly selected for meticulous evaluation to ensure alignment between images and masks. Subsequently, the dataset was divided into training, validation, and test sets using an 80/10/10 split. The UNet model was trained until it achieved the lowest validation loss to optimize performance. During training, the model accurately identified the tumor region in many images but struggled to delineate it accurately, likely due to the dataset's lack of a well-defined pattern. Test results showed an average accuracy, precision, recall, dice, and IoU of 0.86, 0.70, 0.52, 0.52, and 0.38, respectively. Preliminary results demonstrate great potential to apply image segmentation to detect salivary gland tumors in clinical photographs. To enhance results, subsequent steps may involve increasing the dataset size, utilizing data augmentation, preprocessing images (intensity normalization, histogram equalization, noise removal, and contrast enhancement), tuning UNet hyperparameters, conducting error analysis to identify patterns, and filtering out highly dissimilar images. [ABSTRACT FROM AUTHOR]

Early detection of oral squamous cell carcinoma (OSCC) is crucial for reducing its high mortality rates. Deep learning (DL) artificial models have shown promise in recognizing patterns in complex samples, which can be valuable for improving the diagnosis of malignant and potentially malignant oral lesions. To use the convolutional neural network (CNN) deep learning algorithms to develop an automated classification model for OSCC and oral potentially malignant disorders (OPMD) recognition using clinical images, and to assess interpretability of the model's regions of interest through Grad-CAM visual representation. A cross-sectional study was performed retrospectively including a dataset of 837 clinical images obtained from 3 Brazilian oral medicine services, consisting of 351 images of OPMD and 486 images of OSCC. The dataset was divided into training, validation, and test subsets with an 8:1:1 proportion. Two transfer learning strategies were applied, using frozen and fine-tuned weights of ImageNet ILSVRC and ISIC 2019 to initially pre-train the ConvNeXt convolutional neural network (CNN). Pretraining with ILSVRC yielded the best performance for the ConvNeXt CNN model, achieving a mean accuracy, precision, recall, and F1 score of 1.00 during training and 0.78 during testing. Further pretraining with ISIC 2019 resulted in slightly lower training scores (0.99), yet the test scores were comparable to those of the latter strategy. The ConvNeXt CNN model demonstrated high performance in distinguishing OPMD and OSCC. Interpretability of obtained results was facilitated by implementing Grad-CAM assisted representation. [ABSTRACT FROM AUTHOR]

Polymorphous adenocarcinoma (PAC) is a malignant neoplasm of the salivary gland with a low incidence worldwide, characterized by diverse morphological features. This study aimed to present a large and well-documented case series from a single Brazilian center. We analyzed demographic data, clinical presentations, histopathological features, and immunohistochemical characteristics of 26 cases of PAC. Most patients were women (n = 21), with a mean age of 58 years (range = 36 to 84 years). The typical clinical presentation was a fibrous/fibroelastic and firm nodular lesion, with an average diameter of 2 cm, most found on the palate (n = 15). The alveolar ridge, buccal mucous, and superior lip were also affected. Histologically, various patterns were observed, including tubular, solid, cribriform, papillary, and cystic. Glomeruloid slit-like structures (n = 11), clear cells (n = 13), and mucous cells (n = 8) were also identified. Surface papillary epithelial hyperplasia was observed in four cases, and the presence of myxoid and collagenous areas in the stroma in nine. Fusiform areas were noted in 2 cases and areas of squamous differentiation in one. A clear cell predominance was observed in 2 cases. Neural invasion was present in eight cases. All cases were positive for S-100 and Cytokeratin 7 (CK7), and negative for p-40. The Ki-67 index was low, with a mean of 2.5%. In this case series, PAC primarily affected minor salivary glands and exhibited some uncommon microscopic features. Oral pathologists should be aware of these findings to prevent misdiagnosis. [ABSTRACT FROM AUTHOR]

Lately, there has been a great search for molecular-based treatments. The aim of this study was to characterize the salivary proteomic profile of patients treated for oral squamous cell carcinoma (OSCC) and its correlation with the risk of developing severe radiation-related oral toxicities. Thirty-five OSCC patients submitted to radiotherapy (RT) or chemoradiotherapy (CRT) were included. Xerostomia, dysphagia, dysphagia-related pain (DRP), dysgeusia, and oral candidiasis (OC) were evaluated daily. Whole saliva was collected prior to RT and subjected to target proteomic analysis; results were statistically compared to oral toxicities and clinical outcomes. Eighty percent of patients presented stage III/IV OSCC. Sixty-three percent were submitted to CRT protocols, mean RT dose of 66.7 Gy. 42.9% and 60% of patients presented with severe (grades 2-3) xerostomia and dysphagia, respectively; 35.3% presented severe DRP. 68.6% presented grade 2 dysgeusia, and 25.7% presented OC over 4 weeks during RT. Target proteomic analysis revealed a total of 56 proteins from which statical significance was observed in 11 correlated with severe xerostomia, 1 with severe dysphagia, 4 with severe DRP, 1 with dysgeusia, and 19 with over 4 weeks of OC. Eight proteins were concomitant to xerostomia and OC, and 1 protein was concomitant for dysgeusia and OC. The present study is a pioneer in characterizing possible biomarkers that may allow the identification of patients that are more likely to develop severe RT oral toxicities. Further studies are necessary in order to validate and better understand the role of these proteins in the pathophysiology of radiation-related oral toxicities. [ABSTRACT FROM AUTHOR]

Head and neck radiotherapy can cause several toxicities, and its management has important treatment implications. Proper information about treatment is crucial to assist patients by preparing them and enhancing their ability to manage their illness. This study aimed to verify the impact of an educational video on the improvement of the patient's understanding, satisfaction, and quality of life and the influence on their emotional state at different times during treatment. A 10-minute video about head and neck radiotherapy and its toxicities was produced. A prospective randomized clinical trial was performed in 2 groups: the control group (n = 65), which received standard verbal and written information, and the experimental group (n = 65), which received standard information and the video. Appropriate questionnaires (Hospital Anxiety and Depression Scale, University of Washington Quality of Life version 4, IRTU, and Post-RTU) were given at 4 different times to evaluate patients' understanding, anxiety, depression, and quality of life. The video improved the understanding of treatment and its side effects. In addition, the video group reported better awareness about oral health care during treatment. Osteoradionecrosis and radiation-related caries were the most common unknown adverse effects. On the other hand, the educational video did not modify the patients' anxiety, depression, and quality of life. All patients reported high satisfaction with the video. Audio-visual tools may improve patients' understanding of radiotherapy and showed to be a useful tool to be used in association with verbal and written information in cancer centers. In addition, information about osteoradionecrosis and radiation-related caries must be provided to patients. [ABSTRACT FROM AUTHOR]

Photobiomodulation therapy (PBMT) has increasingly been used to manage oral toxicities secondary to cancer treatment; however, its impact on the stimulation of malignant cells is currently unknown. This 4-arm, double-blind, phase III randomized controlled clinical trial (RBR-4w4wwx) was designed to assess the impact of prophylactic PBMT for oral and oropharyngeal mucositis (OM) in the survival rates of patients with oral and oropharyngeal squamous cell carcinoma (OPSCC). Patients were randomly allocated to 4 groups: intraoral prophylactic PBMT (G1), placebo intraoral PBMT (G2), extraoral prophylactic PBMT (G3), and placebo extraoral PBMT (G4). OM grade (NCI) and OM-associated pain (EVA scale) were assessed weekly during OPSCC treatment. Tumor outcomes were evaluated quarterly during 12 months. Data analysis was performed using descriptive statistics. A total of 45 OPSCC patients were included. OM peak ulcerations (grade 2 or higher) were observed at week 3 for G1, G2, and G4 and at week 5 for G3. Placebo groups had the highest levels of pain when compared with the treatment/intervention groups (P <.05). No evidence of oral premalignant lesions or second primary tumors were observed in any of the patients included. In terms of cancer treatment outcomes, tumor persistence and regional metastases were higher among placebo patients after 12 months of follow-up (P <.05). Death due to disease progression was observed in 25% of patients from the treatment/intervention groups and 37.5% of placebo patients (P <.05). No negative impact of prophylactic PBMT for OM could be detected in terms OPSCC treatment outcomes. [ABSTRACT FROM AUTHOR]

Photobiomodulation therapy (PBMT) is recommended for the prevention of radiation-induced oral mucositis (OM) in head and neck patients. It is unclear to what extent the delivered PBMT therapy effectively provides an adequate uniform dose to the at-risk tissues. To characterize oral sites affected by radiation-induced OM and its clinicopathologic outcomes in oral squamous cell carcinoma (OSCC) patients subjected to prophylactic PBMT. The site distribution of OM, OM grading (NCI, Version 4.0), OM-related pain (visual analog scale), analgesic protocol (WHO), and use of enteral nutrition were evaluated weekly during treatment for OSCC patients with prophylactic PBMT. Data analysis was performed using descriptive statistics. One hundred forty-five OSCC patients were included. The most frequently affected sites by OM included lateral tongue (44.1%) and buccal mucosa (37.2%). Keratinized oral mucosae sites, including the tongue dorsum (6.21%) and retromolar trigone (8.3%), were also affected. Peak OM scores were observed at weeks 5, 6 and 7, with severe OM (NCI grades 3 and 4) rates of 11%, 20%, and 25%, respectively. The cumulative occurrence of severe OM was 23%, which developed as early as week 3 and as late as week 7. The highest mean value of OM-related pain (2.7) was observed in the sixth week, and 13.8% of the patients required feeding support. Compared to previously published studies that did not provide PBMT, OM, pain and analgesia and tube feeding use reduction was seen. Keratinized surfaces should also be included in the prophylactic PBMT for OM aiming to reduce severe OM. [ABSTRACT FROM AUTHOR]