A high-order focus interaction model and oral ulcer dataset for oral ulcer segmentation

Association between non-healing precancerous oral lesions and ulcers with tobacco smoking: A population-based study

The automatic segmentation and classification of skin lesions are two essential tasks in computer-aided skin cancer diagnosis. Segmentation aims to detect the location and boundary of the skin lesion area, while classification is used to evaluate the type of skin lesion. The location and contour information of lesions provided by segmentation is essential for the classification of skin lesions, while the skin disease classification helps generate target localization maps to assist the segmentation task. Although the segmentation and classification are studied independently in most cases, we find meaningful information can be explored using the correlation of dermatological segmentation and classification tasks, especially when the sample data are insufficient. In this paper, we propose a collaborative learning deep convolutional neural networks (CL-DCNN) model based on the teacher–student learning method for dermatological segmentation and classification. To generate high-quality pseudo-labels, we provide a self-training method. The segmentation network is selectively retrained through classification network screening pseudo-labels. Specially, we obtain high-quality pseudo-labels for the segmentation network by providing a reliability measure method. We also employ class activation maps to improve the location ability of the segmentation network. Furthermore, we provide the lesion contour information by using the lesion segmentation masks to improve the recognition ability of the classification network. Experiments are carried on the ISIC 2017 and ISIC Archive datasets. The CL-DCNN model achieved a Jaccard of 79.1% on the skin lesion segmentation task and an average AUC of 93.7% on the skin disease classification task, which is superior to the advanced skin lesion segmentation methods and classification methods.

Effect of cisplatin on oral ulcer-induced nociception in rats

There have been studies on the application of computer-aided diagnosis (CAD) in the endoscopic diagnosis of early esophageal cancer (EEC), but there is still a significant gap from clinical application. We developed an endoscopic CAD system for EEC based on the AutoGluon framework, aiming to explore the feasibility of automatic deep learning (DL) in clinical application. The endoscopic pictures of normal esophagus, esophagitis, and EEC were collected from The First Affiliated Hospital of Soochow University (September 2015 to December 2021) and the Norwegian HyperKvasir database. All images of non-cancerous esophageal lesions and EEC in this study were pathologically examined. There were three tasks: task A was normal vs. lesion classification under non-magnifying endoscopy (n=932 vs. 1,092); task B was non-cancer lesion vs. EEC classification under non-magnifying endoscopy (n=594 vs. 429); and task C was non-cancer lesion vs. EEC classification under magnifying endoscopy (n=505 vs. 824). In all classification tasks, we took 100 pictures as the verification set, and the rest comprised as the training set. The CAD system was established based on the AutoGluon framework. Diagnostic performance of the model was compared with that of endoscopists grouped according to years of experience (senior >15 years; junior <5 years). Model evaluation indicators included accuracy, recall rate, precision, F1 value, interpretation time, and the area under the receiver operating characteristic (ROC) curve (AUC). In tasks A and B, the accuracies of medium-performance CAD and high-performance CAD were lower than those of junior doctors and senior doctors. In task C, the medium-performance and high-performance CAD accuracies were close to those of junior doctors and senior doctors. The high-performance CAD model outperformed the junior doctors in both task A (0.850 vs. 0.830) and task C (0.840 vs. 0.830) in sensitivity comparison, but there was still a large gap between high-performance CAD models and doctors in sensitivity comparison. In task A, with the aid of CAD pre-interpretation, the accuracy of junior and senior physicians were significantly improved (from 0.880 to 0.915 and from 0.920 to 0.945, respectively); the time spent on film reading was significantly shortened (junior: from 11.3 to 8.7 s; senior: from 6.7 to 5.5 s). In task C, with the aid of CAD pre-interpretation, the accuracy of junior and senior physicians were significantly improved (from 0.850 to 0.865 and from 0.915 to 0.935, respectively); the reading time was significantly shortened (junior: from 9.5 to 7.7 s; senior: from 5.6 to 3.0 s). The CAD system based on the AutoGluon framework can assist doctors to improve the diagnostic accuracy and reading time of EEC under endoscopy. This study reveals that automatic DL methods are promising in clinical application.

Microvesicles (MVs) are membrane vesicles secreted by cells and are present in the saliva of healthy individuals. It has various functions and has been reported to be a biomarker for malignant tumors. The changes in saliva levels of MVs associated with disease(s) is unclear. This study aimed to determine the proportion of salivary apoptotic MVs and their association with oral ulcer(s) in patients with non-healing oral ulcer(s) and reported oral cancer. Saliva (5 mL) was collected from patients with non-healing oral ulcer(s) and reported oral cancer (at the time of saliva collection, the participant have an oral ulcer(s) in the oral cavity and have an oral cancer lesion; n = 73) and healthy volunteers with oral ulcer(s) (n = 62). A standard differential centrifugation protocol was used for the purification of MVs. Dynamic light scattering and transmission electron microscopy were used to characterize MVs. Flow cytometry was used to quantify salivary apoptotic MVs. Immunocytochemistry was performed according to a standard protocol. None of patients with oral cancer has smoking and drinking habit. The majority of saliva samples derived from patients with non-healing oral ulcer(s) and reported oral cancer were more positive for the fluorescent dye carboxyfluorescein succinimidyl ester than those of healthy volunteers with oral ulcer(s). Salivary fluid obtained from patients had membrane-limited vesicles that were round and/or slightly elongated in shape, with diameters of 100–1,000 nm. The number of salivary apoptotic MVs was higher in patients with non-healing oral ulcer(s) than in those derived from healthy volunteers with oral ulcer(s) (p < 0.001). There was an association between salivary apoptotic MVs in patients with non-healing oral ulcer(s) and the degree or severity of oral ulcers (p < 0.001). Levels of salivary apoptotic MVs are elevated in patients with non-healing oral ulcer(s) and confirmed oral cancer. Elevated levels of salivary apoptotic MVs are associated with clinicopathological data of patients with oral cancer. Evidence level: IV. Technical efficacy: stage 3.

Oral squamous cell carcinoma develops continuously out of predamaged oral mucosa. For the physician and pathologist, difficulties arise in distinguishing precancerous from cancerous lesions. MAGE-A antigens are tumor antigens that are found solely in malignant transformed cells. These antigens might be useful in distinguishing precancerous from cancerous lesions. The aim of this study was to verify this assumption by comparing MAGE-A expression in benign, precancerous, and cancerous lesions of the oral mucosa. Retrospectively, biopsies of different oral lesions were randomly selected. The lesions that were included are 64 benign oral lesions (25 traumatic lesions (oral ulcers), 13 dental follicles, and 26 epulis), 26 oral lichen planus, 123 epithelial precursor lesions (32 epithelial hyperplasia found in leukoplakias, 24 epithelial dysplasia found in leukoplakias, 26 erythroplasia with oral epithelial dysplasia, and 41 carcinomas in situ in erythroleukoplakias). The lesions were immunohistochemically stained with the poly-MAGE-A antibody 57B, and the results were compared. Biopsies of oral lichen planus, oral ulcers, dental follicles, epulis, and leukoplakia without dysplasia showed no positive staining for MAGE-A antigens. Leukoplakia with dysplasia, dysplasia, and carcinomata in situ displayed positive staining in 33%, 65%, and 56% of the cases, respectively. MAGE-A antigens were not detectable via immunohistochemistry in benign lesions of the oral mucosa. The staining rate of dysplastic precancerous lesions or malignant lesions ranged from 33% to 65%. The MAGE-A antigens might facilitate better differentiation between precancerous and cancerous lesions of the oral mucosa.

Background and Aims:In Taiwan, approximately 90% of patients with end-stage renal disease receive maintenance hemodialysis. Although studies have reported the survival predictability of multiclinical factors, the higher-order interactions among these factors have rarely been discussed. Conventional statistical approaches such as regression analysis are inadequate for detecting higher-order interactions. Therefore, this study integrated receiver operating characteristic, logistic regression, and balancing functions for adjusting the ratio in risk classes and classification errors for imbalanced cases and controls using multifactor-dimensionality reduction (MDR-ER) analyses to examine the impact of interaction effects between multiclinical factors on overall mortality in patients on maintenance hemodialysis.Meterials and Methods:In total, 781 patients who received outpatient hemodialysis dialysis three times per week before 1 January 2009 were included; their baseline clinical factor and mortality outcome data were retrospectively collected using an approved data protocol (201800595B0).Results:Consistent with conventional statistical approaches, the higher-order interaction model could indicate the impact of potential risk combination unique to patients on maintenance hemodialysis on the survival outcome, as described previously. Moreover, the MDR-based higher-order interaction model facilitated higher-order interaction effect detection among multiclinical factors and could determine more detailed mortality risk characteristics combinations.Conclusion:Therefore, higher-order clinical risk interaction analysis is a reasonable strategy for detecting non-traditional risk factor interaction effects on survival outcome unique to patients on maintenance hemodialysis and thus clinically achieving whole-scale patient care.

In many patients with cancer, chemotherapy-induced severe oral ulcerative mucositis causes intractable pain, leading to delays and interruptions in therapy. However, the pain mechanism in oral ulcerative mucositis after chemotherapy has not been extensively studied. In this study, we investigated spontaneous pain and mechanical allodynia in a preclinical model of oral ulcerative mucositis after systemic administration of the chemotherapy drug 5-fluorouracil, using our proprietary pain assay system for conscious rats. 5-Fluorouracil caused leukopenia but did not induce pain-related behaviors. After 5-fluorouracil administration, oral ulcers were developed with topical acetic acid treatment. Compared with saline-treated rats, 5-fluorouracil-exposed rats showed more severe mucositis with excessive bacterial loading due to a lack of leukocyte infiltration, as well as enhancements of spontaneous pain and mechanical allodynia. Antibacterial drugs, the lipid A inhibitor polymyxin B and the TRPV1/TRPA1 channel pore-passing anesthetic QX-314, suppressed both the spontaneous pain and the mechanical allodynia. The cyclooxygenase inhibitor indomethacin and the TRPV1 antagonist SB-366791 inhibited the spontaneous pain, but not the mechanical allodynia. In contrast, the TRPA1 antagonist HC-030031 and the N-formylmethionine receptor FPR1 antagonist Boc MLF primarily suppressed the mechanical allodynia. These results suggest that 5-fluorouracil-associated leukopenia allows excessive oral bacterial infection in the oral ulcerative region, resulting in the enhancement of spontaneous pain through continuous TRPV1 activation and cyclooxygenase pathway, and mechanical allodynia through mechanical sensitization of TRPA1 caused by neuronal effects of bacterial toxins. These distinct pain mechanisms explain the difficulties encountered with general treatments for oral ulcerative mucositis-induced pain in patients with cancer and suggest more effective approaches.

Introduction: The oral cavity, apart from being the beginning of the gastrointestinal tract and playing a role in the speech, can also be a basis for symptoms, connected with systemic diseases. The most serious problems concerning dental professionals include caries, oral ulcers, mucosal erythema, gingival bleeding and hypertrophy, soft tissue exophytic masses, dry mouth, facial pain, movement disorders, tooth abnormalities, abnormal dental wear, tooth/mucosal discoloration, developmental and bone pathology. Materials and Methods: Oral manifestations of systemic diseases can be evaluated through a variety of methods. Careful examination of the oral cavity is still the most important procedure. It may reveal mucosal changes, periodontal inflammation and bleeding, and the general condition of the teeth. Biopsy is an obligatory method for diagnosing cancer and precancerous lesions. Even then, there is a possibility for a mistake. The function of oral organs is assessed as a malfunction may indicate an unnoticed until now health problem. Pain symptoms should never be underestimated. Results: Dental specialists should have in mind that even the smallest irregularity may be connected to an undiagnosed systemic disease. Oral ulceration may be found in patients with lupus erythematosus, pemphigus vulgaris, or Crohn`s disease. Severe periodontal inflammation or bleeding should prompt investigation of conditions such as diabetes mellitus, human immunodeficiency virus infection, thrombocytopenia, and leukemia. An understanding of the relationship between systemic disease and oral pathology is important with respect to establishing the diagnosis and determining the complexity of the following treatment. Conclusion: Careful examination of the oral cavity may reveal findings indicative of an underlying systemic condition, and allow for early diagnosis and treatment. All fields of medicine should work together in the best interest of patients.

Objectives: Oral ulceration forms a major category of oral lesions. Due to its effect on the quality of life, the patients seek treatment using various substances. This study investigated the types of treatments used by a sample of patients having all types of oral ulcers and its positive and negative effects on the patients' lives. Methods: This cross-sectional study included patients having different types of oral ulcers. Participants provided information about their previous treatments, the source of recommendation of its use and side effects linked to them. They filled the oral health impact profile-14 questionnaire to report the change in their quality of life after treatment. The patients' educational and social levels were tested as predictors for the use of non-conventional treatments. Results: Correct pharmaceutical treatments were administered y 34% of the participants; the most effective of which were systemic and topical steroids by ulcerative oral lichen planus patients; and vitamin supplements in recurrent aphthous stomatitis patients. The majority (64.2%) of the previously treated patients used non-conventional treatments, the most widely used of which was Tahini. Half of the patients using non-conventional treatments suffered side effects. The source of information of the self-medication was attributed mainly to the patients themselves, followed by friends and family members. Conclusion: This study reflects the widely spread habit of self-medication in our sample of patients having oral ulcers. It highlights the negative influence of these remedies.

Early detection and treatment are essential to improve the prognosis of gastric cancer. Screening endoscopy plays an important role in early gastric cancer (EGC) detection.1 Endoscopic diagnosis of EGC consists of two steps, detection and characterization, typically with a targeted biopsy. During screening gastroscopy, clinicians generally perform endoscopic detection of EGC with white light endoscopy based on subtle color or morphological change, such as reddish or pale elevation or depression, compared with the surrounding area. The following step is endoscopic lesion characterization, the differentiation between noncancerous and cancerous lesions, based on lesion morphology, for example, irregular margin and spiny depressed area.2 However, sufficient training is required to achieve a satisfactory endoscopic diagnosis of EGC during screening endoscopy. Artificial intelligence (AI) has recently made remarkable progress, and computer-aided detection (CADe) and computer-aided diagnosis (CADx) systems for gastrointestinal endoscopy are also rapidly growing.3 This research area is expected to provide the solution for several limitations of current clinical practice. Some researchers developed the CADe system for gastric cancer detection and reported a favorable detection performance.4, 5 In light of the background, Ishioka et al.6 developed a CADx to differentiate non-neoplastic from neoplastic gastric lesions in white light endoscopy, named Tango. The authors evaluated the differential diagnostic performance of Tango compared with specialists and nonspecialists, using still images of white light endoscopy. The sensitivity, the primary outcome of the present study, was 84.7% for Tango, 65.8% for the specialists, and 51.0% for nonspecialists, demonstrating the superiority of Tango over the specialists. Also, the authors showed the noninferiority of Tango to specialists in the accuracy (70.8% vs. 67.4%), and the superiority over nonspecialists in the sensitivity and the accuracy (84.7% vs. 51.0% and 70.8% vs. 58.4%, respectively). As expected, the time to diagnosis was significantly shorter in Tango (0.08 s/image) than in the endoscopists (over 4 s/image). The present study highlighted that Tango allows for sensitive and fast diagnostic performance to differentiate between non-neoplastic and neoplastic lesions. The concept of the present study is valuable because Tango is a promising support tool that could avoid misdiagnosis during gastric cancer screening and educate nonexperienced endoscopists. Nevertheless, we should interpret the data and the results carefully, particularly when utilizing the AI support tool in real clinical practice. First, the authors developed Tango on the premise of appropriate lesion detection. In other words, Tango can work well only on still images that are appropriately captured from a suitable distance in real clinical practice. Please be reminded of the first challenging step, the endoscopic detection of EGC. Clinicians should adopt the CADe system in combination with a CADx system, which allows for endoscopic detection and characterization at once during screening endoscopy. All endoscopists require the skill to perform the basic systematic stomach protocol to inspect the entire stomach and avoid missed gastric cancer, which is recommended by the Japanese guidelines.7 In addition, the former studies using the CADe systems also showed sensitivities and specificities. Researchers should clarify the difference in the data interpretation. Second, the specificity of Tango was lower than that of nonspecialists and specialists. Although researchers should consider the sensitivity more critical than the specificity not to miss EGC, the trade-off may be one of the commonly raised issues to be resolved when utilizing the AI technology in clinical practice because it can cause an increase in unnecessary biopsies.8 Also, this AI technology will not work for nonexpert endoscopists' training in diagnosing non-neoplastic lesions because the specificity is lower in Tango than in nonexpert endoscopists. Third, the guidelines recommend using image-enhanced endoscopy for the qualitative diagnosis of EGC.7 Specifically, magnifying endoscopy with narrow-band imaging (NBI) offers excellent diagnostic yield to differentiate cancerous and noncancerous lesions based on magnifying endoscopy simple diagnostic algorithm for early gastric cancer.9 The algorithm can be well adopted into the CADe. Ueyama et al. developed a CADx to differentiate noncancerous and cancerous lesions based on magnifying endoscopy with NBI and reported excellent diagnostic performance with overall accuracy, sensitivity, and specificity of 98.7%, 98%, and 100%, respectively.10 However, to utilize this AI system in real clinical practice, it is essential to perform high-quality magnifying endoscopy with NBI, which requires a certain amount of training. The excellent diagnostic performance of Ueyama et al. is thought to be established based on a perfect quality full zoom magnification. In addition, magnifying endoscopy is not always available, especially in screening endoscopy. Thus, Tango has the potential to be an easy-to-use good alternative to CADe based on magnifying endoscopy with NBI. However, as mentioned earlier, all CADe systems are developed on the premise of appropriate lesion detection, which is an unavoidable challenge in gastric cancer screening. The current limitations and weaknesses should be resolved by further development of AI technology and improving the algorithm for the combined use of CADe and CADx, and the dataset for convolutional neural network training by using a larger sample size obtained from videos. Further large-scale prospective studies are desirable to confirm the efficacy of Tango. Also, I hope that CADe and CADx will work together as if dancing the tango to support and contribute to a sharp endoscopic diagnosis of gastric cancer. Author declares no conflict of interest for this article. None.

Artificial Intelligence (AI) has achieved remarkable performance in the field of medical image analysis, particularly in tasks such as object detection, segmentation, and classification. In this paper, we introduce a solution for automatic breast cancer diagnosis based on the U-Net architecture, which we call (U-Net)+. The novel (U-Net)+ is designed to handle both segmentation and classification tasks within a signal framework. We retained the original U-Net architecture due to its strong learning capabilities and its advantages in semantic segmentation. Notable, we incorporated fully connected layers into the bottleneck layers, serving as a multi-functional classifier for both initial diagnoses based on raw images and further diagnoses for segmented images. The (U-Net)+ model is trained using a joint loss function. We conducted the experiments on breast ultrasound images, demonstrating that the (U-Net) performs well in both classification and segmentation tasks

Learning from dermoscopic images in association with clinical metadata for skin lesion segmentation and classification

Accurate segmentation of leaf diseases is crucial for crop health management and disease prevention. However, existing studies fall short in addressing issues such as blurred disease spot boundaries and complex feature distributions in disease images. Although the vision foundation model, Segment Anything Model (SAM), performs well in general segmentation tasks within natural scenes, it does not exhibit good performance in plant disease segmentation. To achieve fine-grained segmentation of leaf disease images, this study proposes an advanced model: Enhanced Multi-Scale SAM (EMSAM). EMSAM employs the Local Feature Extraction Module (LFEM) and the Global Feature Extraction Module (GFEM) to extract local and global features from images respectively. The LFEM utilizes multiple convolutional layers to capture lesion boundaries and detailed characteristics, while the GFEM fine-tunes ViT blocks using a Multi-Scale Adaptive Adapter (MAA) to obtain multi-scale global information. Both outputs of LFEM and GFEM are then effectively fused in the Feature Fusion Module (FFM), which is optimized with cross-branch and channel attention mechanisms, significantly enhancing the model's ability to handle blurred boundaries and complex shapes. EMSAM integrates lightweight linear layers as classification heads and employs a joint loss function for both classification and segmentation tasks. Experimental results on the PlantVillage dataset demonstrate that EMSAM outperforms the second-best state-of-the-art semantic segmentation model by 2.45% in Dice Coefficient and 6.91% in IoU score, and surpasses the baseline method by 21.40% and 22.57%, respectively. Particularly, for images with moderate and severe disease levels, EMSAM achieved Dice Coefficients of 0.8354 and 0.8178, respectively, significantly outperforming other semantic segmentation algorithms. Additionally, the model achieved a classification accuracy of 87.86% across the entire dataset, highlighting EMSAM's effectiveness and superiority in plant disease segmentation and classification tasks.

White matter hyperintensities (WMH) represent neuroimaging markers of cerebral small vessel disease. We aimed to develop and validate a deep learning-based simultaneous, automatic WMH segmentation and classification model in patients with cognitive impairment. This retrospective study included images from consecutive patients with cognitive impairment from a tertiary hospital. A segmentation model was trained and tuned on 448 and 149 subjects. A classification compartment inherited from the segmentation model, utilizing the multi-task multiple instance learning framework (MTMIL), was trained and tuned on 1186 and 394 subjects. Tests of segmentation and classification tasks were performed on 149 and 394 subjects in the internal testing dataset and 100 subjects in the external dataset. We evaluated five different models to select the segmentation branch.Classification according to the Fazekas scale used three categories (normal/mild, moderate, severe), and locations were separately reported as periventricular and deep WMH. For classification evaluation, three experienced neuroradiologists evaluated test datasets. Between January 2016 and December 2019, 1974 consecutive patients (mean age 71.1 ± 9.7) were included. Dice score performance of the UNet with Resnet-34 encoder model on internal and external testing datasets was 0.88 (95% CI: 0.88-0.89) and 0.85 (95% CI: 0.84-0.86) for single-class segmentation, and 0.77 (95% CI: 0.76-0.79) and 0.72 (95% CI: 0.71-0.74) for multi-class segmentation. The accuracy of the Fazekas scale classification model was 0.88 and 0.87 at periventricular and deep WMH with internal datasets and 0.68 and 0.75 with external datasets. These results demonstrate the high segmentation and classification performance of our models and their potential for deployment as accurate diagnostic support tools for quantified evaluation of WMH. Question We developed a deep learning-based, simultaneous, automatic white matter hyperintensity (WMH) segmentation and classification model using data from patients with cognitive impairment. Findings Segmentation model achieved Dice scores of 0.72-0.88 for single-class and multi-class segmentation. Fazekas score classification accuracy ranged from 0.68 to 0.88 for periventricular/deep WMH. Clinical relevance Our study demonstrated high segmentation and classification performance of deep learning-based models and potential for deployment as accurate diagnostic support tools for quantified evaluation of white matter hyperintensities.