Search

Convolutional neural networks have demonstrated excellent performance in oral cancer detection and classification. However, the end-to-end learning strategy makes CNNs hard to interpret, and it can be challenging to fully understand the decision-making procedure. Additionally, reliability is also a significant challenge for CNN based approaches. In this study, we proposed a neural network called the attention branch network (ABN), which combines the visual explanation and attention mechanisms to improve the recognition performance and interpret the decision-making simultaneously. We also embedded expert knowledge into the network by having human experts manually edit the attention maps for the attention mechanism. Our experiments have shown that ABN performs better than the original baseline network. By introducing the Squeeze-and-Excitation (SE) blocks to the network, the cross-validation accuracy increased further. Furthermore, we observed that some previously misclassified cases were correctly recognized after updating by manually editing the attention maps. The cross-validation accuracy increased from 0.846 to 0.875 with the ABN (Resnet18 as baseline), 0.877 with SE-ABN, and 0.903 after embedding expert knowledge. The proposed method provides an accurate, interpretable, and reliable oral cancer computer-aided diagnosis system through visual explanation, attention mechanisms, and expert knowledge embedding.

SignificanceOral cancer is one of the most prevalent cancers, especially in middle- and low-income countries such as India. Automatic segmentation of oral cancer images can improve the diagnostic workflow, which is a significant task in oral cancer image analysis. Despite the remarkable success of deep-learning networks in medical segmentation, they rarely provide uncertainty quantification for their output.AimWe aim to estimate uncertainty in a deep-learning approach to semantic segmentation of oral cancer images and to improve the accuracy and reliability of predictions.ApproachThis work introduced a UNet-based Bayesian deep-learning (BDL) model to segment potentially malignant and malignant lesion areas in the oral cavity. The model can quantify uncertainty in predictions. We also developed an efficient model that increased the inference speed, which is almost six times smaller and two times faster (inference speed) than the original UNet. The dataset in this study was collected using our customized screening platform and was annotated by oral oncology specialists.ResultsThe proposed approach achieved good segmentation performance as well as good uncertainty estimation performance. In the experiments, we observed an improvement in pixel accuracy and mean intersection over union by removing uncertain pixels. This result reflects that the model provided less accurate predictions in uncertain areas that may need more attention and further inspection. The experiments also showed that with some performance compromises, the efficient model reduced computation time and model size, which expands the potential for implementation on portable devices used in resource-limited settings.ConclusionsOur study demonstrates the UNet-based BDL model not only can perform potentially malignant and malignant oral lesion segmentation, but also can provide informative pixel-level uncertainty estimation. With this extra uncertainty information, the accuracy and reliability of the model’s prediction can be improved.

Early detection of oral cancer in low-resource settings necessitates a Point-of-Care screening tool that empowers Frontline-Health-Workers (FHW). This study was conducted to validate the accuracy of Convolutional-Neural-Network (CNN) enabled m(mobile)-Health device deployed with FHWs for delineation of suspicious oral lesions (malignant/potentially-malignant disorders). The effectiveness of the device was tested in tertiary-care hospitals and low-resource settings in India. The subjects were screened independently, either by FHWs alone or along with specialists. All the subjects were also remotely evaluated by oral cancer specialist/s. The program screened 5025 subjects (Images: 32,128) with 95% (n = 4728) having telediagnosis. Among the 16% (n = 752) assessed by onsite specialists, 20% (n = 102) underwent biopsy. Simple and complex CNN were integrated into the mobile phone and cloud respectively. The onsite specialist diagnosis showed a high sensitivity (94%), when compared to histology, while telediagnosis showed high accuracy in comparison with onsite specialists (sensitivity: 95%; specificity: 84%). FHWs, however, when compared with telediagnosis, identified suspicious lesions with less sensitivity (60%). Phone integrated, CNN (MobileNet) accurately delineated lesions (n = 1416; sensitivity: 82%) and Cloud-based CNN (VGG19) had higher accuracy (sensitivity: 87%) with tele-diagnosis as reference standard. The results of the study suggest that an automated mHealth-enabled, dual-image system is a useful triaging tool and empowers FHWs for oral cancer screening in low-resource settings.

SignificanceConvolutional neural networks (CNNs) show the potential for automated classification of different cancer lesions. However, their lack of interpretability and explainability makes CNNs less than understandable. Furthermore, CNNs may incorrectly concentrate on other areas surrounding the salient object, rather than the network's attention focusing directly on the object to be recognized, as the network has no incentive to focus solely on the correct subjects to be detected. This inhibits the reliability of CNNs, especially for biomedical applications.AimDevelop a deep learning training approach that could provide understandability to its predictions and directly guide the network to concentrate its attention and accurately delineate cancerous regions of the image.ApproachWe utilized Selvaraju et al.'s gradient-weighted class activation mapping to inject interpretability and explainability into CNNs. We adopted a two-stage training process with data augmentation techniques and Li et al.'s guided attention inference network (GAIN) to train images captured using our customized mobile oral screening devices. The GAIN architecture consists of three streams of network training: classification stream, attention mining stream, and bounding box stream. By adopting the GAIN training architecture, we jointly optimized the classification and segmentation accuracy of our CNN by treating these attention maps as reliable priors to develop attention maps with more complete and accurate segmentation.ResultsThe network's attention map will help us to actively understand what the network is focusing on and looking at during its decision-making process. The results also show that the proposed method could guide the trained neural network to highlight and focus its attention on the correct lesion areas in the images when making a decision, rather than focusing its attention on relevant yet incorrect regions.ConclusionsWe demonstrate the effectiveness of our approach for more interpretable and reliable oral potentially malignant lesion and malignant lesion classification.

In medical imaging, deep learning-based solutions have achieved state-of-the-art performance. However, reliability restricts the integration of deep learning into practical medical workflows since conventional deep learning frameworks cannot quantitatively assess model uncertainty. In this work, we propose to address this shortcoming by utilizing a Bayesian deep network capable of estimating uncertainty to assess oral cancer image classification reliability. We evaluate the model using a large intraoral cheek mucosa image dataset captured using our customized device from high-risk population to show that meaningful uncertainty information can be produced. In addition, our experiments show improved accuracy by uncertainty-informed referral. The accuracy of retained data reaches roughly 90% when referring either 10% of all cases or referring cases whose uncertainty value is greater than 0.3. The performance can be further improved by referring more patients. The experiments show the model is capable of identifying difficult cases needing further inspection.

SignificanceEarly detection of oral cancer is vital for high-risk patients, and machine learning-based automatic classification is ideal for disease screening. However, current datasets collected from high-risk populations are unbalanced and often have detrimental effects on the performance of classification.AimTo reduce the class bias caused by data imbalance.ApproachWe collected 3851 polarized white light cheek mucosa images using our customized oral cancer screening device. We use weight balancing, data augmentation, undersampling, focal loss, and ensemble methods to improve the neural network performance of oral cancer image classification with the imbalanced multi-class datasets captured from high-risk populations during oral cancer screening in low-resource settings.ResultsBy applying both data-level and algorithm-level approaches to the deep learning training process, the performance of the minority classes, which were difficult to distinguish at the beginning, has been improved. The accuracy of "premalignancy" class is also increased, which is ideal for screening applications.ConclusionsExperimental results show that the class bias induced by imbalanced oral cancer image datasets could be reduced using both data- and algorithm-level methods. Our study may provide an important basis for helping understand the influence of unbalanced datasets on oral cancer deep learning classifiers and how to mitigate.

Non-invasive strategies that can identify oral malignant and dysplastic oral potentially-malignant lesions (OPML) are necessary in cancer screening and long-term surveillance. Optical coherence tomography (OCT) can be a rapid, real time and non-invasive imaging method for frequent patient surveillance. Here, we report the validation of a portable, robust OCT device in 232 patients (lesions: 347) in different clinical settings. The device deployed with algorithm-based automated diagnosis, showed efficacy in delineation of oral benign and normal (n = 151), OPML (n = 121), and malignant lesions (n = 75) in community and tertiary care settings. This study showed that OCT images analyzed by automated image processing algorithm could distinguish the dysplastic-OPML and malignant lesions with a sensitivity of 95% and 93%, respectively. Furthermore, we explored the ability of multiple (n = 14) artificial neural network (ANN) based feature extraction techniques for delineation high grade-OPML (moderate/severe dysplasia). The support vector machine (SVM) model built over ANN, delineated high-grade dysplasia with sensitivity of 83%, which in turn, can be employed to triage patients for tertiary care. The study provides evidence towards the utility of the robust and low-cost OCT instrument as a point-of-care device in resource-constrained settings and the potential clinical application of device in screening and surveillance of oral cancer.

SignificanceOral cancer is among the most common cancers globally, especially in low- and middle-income countries. Early detection is the most effective way to reduce the mortality rate. Deep learning-based cancer image classification models usually need to be hosted on a computing server. However, internet connection is unreliable for screening in low-resource settings.AimTo develop a mobile-based dual-mode image classification method and customized Android application for point-of-care oral cancer detection.ApproachThe dataset used in our study was captured among 5025 patients with our customized dual-modality mobile oral screening devices. We trained an efficient network MobileNet with focal loss and converted the model into TensorFlow Lite format. The finalized lite format model is ∼16.3  MB and ideal for smartphone platform operation. We have developed an Android smartphone application in an easy-to-use format that implements the mobile-based dual-modality image classification approach to distinguish oral potentially malignant and malignant images from normal/benign images.ResultsWe investigated the accuracy and running speed on a cost-effective smartphone computing platform. It takes ∼300  ms to process one image pair with the Moto G5 Android smartphone. We tested the proposed method on a standalone dataset and achieved 81% accuracy for distinguishing normal/benign lesions from clinically suspicious lesions, using a gold standard of clinical impression based on the review of images by oral specialists.ConclusionsOur study demonstrates the effectiveness of a mobile-based approach for oral cancer screening in low-resource settings.

The speed and scale of the global COVID-19 pandemic has resulted in unprecedented pressures on health services worldwide, requiring new methods of service delivery during the health crisis. In the setting of severe resource constraint and high risk of infection to patients and clinicians, there is an urgent need to identify consensus statements on head and neck surgical oncology practice. We completed a modified Delphi consensus process of three rounds with 40 international experts in head and neck cancer surgical, radiation, and medical oncology, representing 35 international professional societies and national clinical trial groups. Endorsed by 39 societies and professional bodies, these consensus practice recommendations aim to decrease inconsistency of practice, reduce uncertainty in care, and provide reassurance for clinicians worldwide for head and neck surgical oncology in the context of the COVID-19 pandemic and in the setting of acute severe resource constraint and high risk of infection to patients and staff.

Oral cancer is a growing health issue in low- and middle-income countries due to betel quid, tobacco, and alcohol use and in younger populations of middle- and high-income communities due to the prevalence of human papillomavirus. The described point-of-care, smartphone-based intraoral probe enables autofluorescence imaging and polarized white light imaging in a compact geometry through the use of a USB-connected camera module. The small size and flexible imaging head improves on previous intraoral probe designs and allows imaging the cheek pockets, tonsils, and base of tongue, the areas of greatest risk for both causes of oral cancer. Cloud-based remote specialist and convolutional neural network clinical diagnosis allow for both remote community and home use. The device is characterized and preliminary field-testing data are shared.

ObjectivesSurgical margin status is a significant determinant of treatment outcome in oral cancer. Negative surgical margins can decrease the loco-regional recurrence by five-fold. The current standard of care of intraoperative clinical examination supplemented by histological frozen section, can result in a risk of positive margins from 5 to 17 percent. In this study, we attempted to assess the utility of intraoperative optical coherence tomography (OCT) imaging with automated diagnostic algorithm to improve on the current method of clinical evaluation of surgical margin in oral cancer.Materials and methodsWe have used a modified handheld OCT device with automated algorithm based diagnostic platform for imaging. Intraoperatively, images of 125 sites were captured from multiple zones around the tumor of oral cancer patients (n = 14) and compared with the clinical and pathologic diagnosis.ResultsOCT showed sensitivity and specificity of 100%, equivalent to histological diagnosis (kappa, ĸ = 0.922), in detection of malignancy within tumor and tumor margin areas. In comparison, for dysplastic lesions, OCT-based detection showed a sensitivity of 92.5% and specificity of 68.8% and a moderate concordance with histopathology diagnosis (ĸ = 0.59). Additionally, the OCT scores could significantly differentiate squamous cell carcinoma (SCC) from dysplastic lesions (mild/moderate/severe; p ≤ 0.005) as well as the latter from the non-dysplastic lesions (p ≤ 0.05).ConclusionThe current challenges associated with clinical examination-based margin assessment could be improved with intra-operative OCT imaging. OCT is capable of identifying microscopic tumor at the surgical margins and demonstrated the feasibility of mapping of field cancerization around the tumor.

Oral cancer is the sixth most common cancer worldwide, predominantly seen in low and middle-income countries (LMIC). Two thirds of all cases are detected at a late stage when prognosis and treatment outcomes are poor. Oral lesions are commonly detected by visual inspection, followed by invasive surgical biopsy and time-consuming histopathological analysis. Optical coherence tomography (OCT), a minimally invasive tomographic imaging technology, can be used to non-invasively identify premalignant or malignant change in the oral mucosa. In this study, a mobile OCT imaging system was designed, constructed, and tested for its performance as a point-of-care oral diagnostic device in an LMIC. 20 patients with suspicious oral lesions and 10 healthy subjects were enrolled in this pilot study. Two-dimensional (2-D) OCT images as well as clinical examination data, risk habit history, and histopathology were collected. OCT images for healthy oral mucosa, dysplasia, and malignancy were evaluated in a blinded fashion by visual scoring and computed image processing techniques. It was found that the OCT image processing algorithm performed at or exceeded the performance of visual observer scoring of OCT images.

With the goal to screen high-risk populations for oral cancer in low- and middle-income countries (LMICs), we have developed a low-cost, portable, easy to use smartphone-based intraoral dual-modality imaging platform. In this paper we present an image classification approach based on autofluorescence and white light images using deep learning methods. The information from the autofluorescence and white light image pair is extracted, calculated, and fused to feed the deep learning neural networks. We have investigated and compared the performance of different convolutional neural networks, transfer learning, and several regularization techniques for oral cancer classification. Our experimental results demonstrate the effectiveness of deep learning methods in classifying dual-modal images for oral cancer detection.

AimThe incidence of oral cancer is high in India, which can be reduced by early detection. We aimed to empower frontline health care providers (FHP) for early detection and connect specialist to rural population through mHealth.Materials and methodsWe provided training to FHPs in examination of oral cavity, use of mobile phone for image capture, and risk factor analysis. The FHPs were selected from different cohorts in resource-constrained settings. The workflow involved screening of high-risk individuals in door-to-door and workplace settings, and capture of images of suspected lesions. Uploaded data were interpreted and recommendation was sent by specialist from a remote location. Their recommendation was intimated to FHPs who arranged for further action. Two more initiatives, one for multiple dental schools and another for private practitioners, were undertaken.ResultsDuring the period from 2010 to 2018, 42,754 subjects have been screened, and 5,406 subjects with potentially malignant disorders have been identified. The prevalence of potentially malignant disorders varied from 0.8 to 62% at different cohorts; 516 biopsies have been performed at remote locations.ConclusionConnecting specialists to rural population was made possible through the use of mobile health. Trained FHP were able to reach out to the population. Electronic data capture facilitated efficient follow-up. The program was very cost-effective with screening completed under $1 per person.Clinical significanceIn view of the high incidence of oral cancer in India, and the resource-constrained settings, mobile health paves the way for better access to specialist care for the rural population.

Oral cancer is a growing health issue in a number of low- and middle-income countries (LMIC), particularly in South and Southeast Asia. The described dual-modality, dual-view, point-of-care oral cancer screening device, developed for high-risk populations in remote regions with limited infrastructure, implements autofluorescence imaging (AFI) and white light imaging (WLI) on a smartphone platform, enabling early detection of pre-cancerous and cancerous lesions in the oral cavity with the potential to reduce morbidity, mortality, and overall healthcare costs. Using a custom Android application, this device synchronizes external light-emitting diode (LED) illumination and image capture for AFI and WLI. Data is uploaded to a cloud server for diagnosis by a remote specialist through a web app, with the ability to transmit triage instructions back to the device and patient. Finally, with the on-site specialist's diagnosis as the gold-standard, the remote specialist and a convolutional neural network (CNN) were able to classify 170 image pairs into 'suspicious' and 'not suspicious' with sensitivities, specificities, positive predictive values, and negative predictive values ranging from 81.25% to 94.94%.

Background: Cancer burden in India is rapidly growing, with oral, breast, and uterine cervix being the three most commonly affected sites. It has a catastrophic epidemiological and financial impact on rural communities, the vast majority of whom are socio‐economically disadvantaged. Strengthening the health system is necessary to address challenges in the access and provision of cancer services, thus improving outcomes among vulnerable populations. Objective: To develop, test, and validate a health system capacity assessment (HSCA) tool that evaluates the capacity and readiness for cancer services provision in rural India. Methods: A multi‐method process was pursued to develop a cancer‐specific HSCA tool. Firstly, item generation entailed both a nominal group technique (to identify the health system dimensions to capture) and a rapid review of published and gray literature (to generate items within each of the selected dimensions). Secondly, tool development included the pre‐testing of questionnaires through healthcare facility visits, and item reduction through a series of in‐depth interviews (IDIs) with key local stakeholders. Thirdly, tool validation was performed through expert consensus. Results: A three‐step HSCA multi‐method tool was developed comprising: (a) desk review template, investigating policies and protocols at the state level, (b) facility assessment protocol and checklist, catering to the Indian public healthcare system, and (c) IDI topic guide, targeting policymakers, healthcare workforce, and other relevant stakeholders. Conclusions: The resulting HSCA tool assesses health system capacity, thus contributing to the planning and implementation of context‐appropriate, sustainable, equity‐focused, and integrated early detection interventions for cancer control, especially toward vulnerable populations in rural India and other low‐resource settings. [ABSTRACT FROM AUTHOR]