Adolescents' Perspective Regarding a Community-Wide Mass Drug Administration Program for Soil-Transmitted Helminths in India.

This study was undertaken to understand the perspective of adolescents in endemic communities of India regarding soil-transmitted helminth (STH) infections and community-wide mass drug administration (cMDA). A multicountry community-based cluster-randomized trial, the Deworm3 trial, tested the feasibility of interrupting STH transmission with cMDA, where all individuals aged 1-99 are treated empirically with albendazole. Using a guideline based on the Consolidated Framework for Implementation Research, eight focus group discussions were conducted among 57 adolescents from the trial site in India and analyzed on ATLAS.ti 8.0 software using an a priori thematic codebook. Adolescents believed that adults could be a source of STH infection because they were not routinely dewormed like the children through the national deworming program. Perceived benefits of cMDA for all were better health and increased work efficiency. Perceived barriers to adults' participation in cMDA was their mistrust about the program, fear of side effects, perceived low risk of infection, and absence during drug distribution. To encourage adult participation in cMDAs, adolescents suggested community outreach activities, engaging village influencers and health workers, and tailoring drug distribution to when adults would be available. Adolescents were confident in their ability to be change agents within their households for treatment compliance. Adolescents provided insights into potential barriers and solutions to improve adult participation in cMDA, identified best practices of cMDA delivery, and suggested that they have unique roles as change agents to increase their household participation in cMDA.

Remote evaluation of STH program coverage: Experiences from the DeWorm3 study, India.

BACKGROUND: The DeWorm3 trial is a multi-country study testing the feasibility of interrupting transmission of soil-transmitted helminths by community-wide mass drug administration (cMDA). Treatment coverage during cMDA delivery was validated by in-person coverage evaluation surveys (CES) after each round of treatment. A mobile phone-based CES was carried out in India when access to households was restricted during the COVID-19 lockdown. METHODS: Two focus group discussions were conducted with the survey implementers to document their experiences of conducting phone-based CES via mobile-phone voice calls. PRINCIPAL FINDINGS: In the phone-based CES, only 56% of sampled households were reached compared to 89% during the in-person CES (89%). This was due to phone numbers being wrongly recorded, or calls being unanswered leading to a higher number of households that had to be sampled in order to achieve the sample size of 2,000 households in phone-based CES compared in-person CES (3,600 and 2,352 respectively). Although the phone-based CES took less time to complete than in person coverage evaluations, the surveyors highlighted the lack of gender representation among phone survey participants as it was mostly men who answered calls and were then interviewed. The surveyors also mentioned that eliciting responses to open-ended questions and confirming treatment compliance from every member of the household was challenging during phone based CES. These observations were confirmed by analysing the survey participation data which showed women's participation in CES was significantly lower in phone-based CES (66%) compared to in-person CES (94%) (Z = -22.38; p<0.01) and that a significantly higher proportion of households provided proxy responses in phone-based CES (51%) compared to in-person CES (21%) (Z = 20.23; p<0.01). CONCLUSIONS: The phone-based CES may be a viable option to evaluate treatment coverage but issues such as participation bias, gender inclusion, and quality of responses will need to be addressed to optimize this methodology.

Inter-observer agreement among specialists in the diagnosis of oral potentially malignant disorders and oral cancer using store-and-forward technology.

OBJECTIVES: Oral cancer is a leading cause of morbidity and mortality. Screening and mobile Health (mHealth)-based approach facilitates early detection remotely in a resource-limited settings. Recent advances in eHealth technology have enabled remote monitoring and triage to detect oral cancer in its early stages. Although studies have been conducted to evaluate the diagnostic efficacy of remote specialists, to our knowledge, no studies have been conducted to evaluate the consistency of remote specialists. The aim of this study was to evaluate interobserver agreement between specialists through telemedicine systems in real-world settings using store-and-forward technology. MATERIALS AND METHODS: The two remote specialists independently diagnosed clinical images (n=822) from image archives. The onsite specialist diagnosed the same participants using conventional visual examination, which was tabulated. The diagnostic accuracy of two remote specialists was compared with that of the onsite specialist. Images that were confirmed histopathologically were compared with the onsite diagnoses and the two remote specialists. RESULTS: There was moderate agreement (k= 0.682) between two remote specialists and (k= 0.629) between the onsite specialist and two remote specialists in the diagnosis of oral lesions. The sensitivity and specificity of remote specialist 1 were 92.7% and 83.3%, respectively, and those of remote specialist 2 were 95.8% and 60%, respectively, each compared with histopathology. CONCLUSION: The diagnostic accuracy of the two remote specialists was optimal, suggesting that "store and forward" technology and telehealth can be an effective tool for triage and monitoring of patients. CLINICAL RELEVANCE: Telemedicine is a good tool for triage and enables faster patient care in real-world settings.

Representativeness of a mobile phone-based coverage evaluation survey following mass drug administration for soil-transmitted helminths: a comparison of participation between two cross-sectional surveys.

OBJECTIVES: With increasing mobile phone subscriptions, phone-based surveys are gaining popularity with public health programmes. Despite advantages, systematic exclusion of participants may limit representativeness. Similar to control programmes for neglected tropical diseases (NTDs), the DeWorm3 trial of biannual community-wide mass drug administration (MDA) for elimination of soil-transmitted helminth infection used in-person coverage evaluation surveys to measure the proportion of the at-risk population treated during MDA. Due to lockdown during the COVID-19 pandemic, a phone-based coverage evaluation survey was necessary, providing an opportunity for the current study to compare representativeness and implementation (including non-response) of these two survey modes. DESIGN: Comparison of two cross-sectional surveys. SETTING: The DeWorm3 trial site in Tamil Nadu, India, includes Timiri, a rural subsite, and Jawadhu Hills, a hilly, hard-to-reach subsite inhabited predominantly by a tribal population. PARTICIPANTS: In the phone-based and in-person coverage evaluation surveys, all individuals residing in 2000 randomly selected households (50 in each of the 40 trial clusters) were eligible to participate. Here, we characterise household participation. RESULTS: Of 2000 households, 1780 (89.0%) participated during the in-person survey. Of 2000 households selected for the phone survey, 346 (17.3%) could not be contacted as they had not provided a telephone number during the census and 1144 (57.2%) participated. Smaller households, households with lower socioeconomic status and those with older, women or less educated household-heads were under-represented in the phone-based survey compared with censused households. Regression analysis revealed non-response in the phone-based survey was higher among households from the poorest socioeconomic quintile (prevalence ratio (PR) 2.3, 95% CI 2.0 to 2.7) and lower when heads of households had completed secondary school or higher education (PR 0.7, 95% CI 0.6 to 0.8). CONCLUSIONS: Our findings suggest phone-based surveys under-represent households likely to be at higher risk of NTDs and in-person surveys are more appropriate for measuring MDA coverage within programmatic settings. TRIAL REGISTRATION NUMBER: NCT03014167.

Perceptions, health seeking behavior and utilization of maternal and newborn health services among an indigenous tribal community in Northeast India-a community-based mixed methods study.

Background: Evidence suggests that healthcare utilization among tribal communities in isolated regions can be influenced by social determinants of health, particularly cultural and geographical factors. The true mortality and morbidity due to these factors in remote tribal communities are often underestimated due to facility-dependent reporting systems often difficult to access. We studied the utilization of health services for maternal and newborn care and explored how cultural beliefs, perceptions, and practices influence the health-seeking behavior (HSB) of an indigenous tribal community in Northeast India. Methods: Within a concurrent triangulation design, the combined results from 7 focus group discussions and 19 in-depth interviews, and the 109 interviews of mothers from a community-based survey were interpreted in a complementary manner. The qualitative data were analyzed using a conceptual framework adapted from the socio-ecological and three-delays model, using a priori thematic coding. Multivariable logistic regression was carried out to identify factors associated with home delivery. Results: Only 3.7% of the interviewed mothers received the four recommended antenatal check-ups in health centers, and 40.1% delivered at home. Mothers residing in the villages without a health center or one that was not operational were more likely to deliver at home. HSB was influenced significantly by available finances, the mother's education, low self-esteem, and a strong belief in traditional medicine favored by its availability and religious affiliation. The community sought health services in facilities only in emergency situations, determined primarily by the tribe's poor perception of the quality of health services provided in the irregularly open centers, locally available traditional medicine practitioners, and challenges in geographical access. National schemes intended to incentivize access to facilities failed to impact this community due to flawed program implementation that did not consider this region's cultural, social, and geographical differences. Conclusion: The health-seeking behavior of the tribe is a complex, interrelated, and interdependent process framed in a medical pluralistic context. The utilization of health centers and HSBs of indigenous communities may improve when policymakers adopt a "bottom-up approach," addressing structural barriers, tailoring programs to be culturally appropriate, and guaranteeing that the perceived needs of indigenous communities are met before national objectives.

Inter-observer agreement among specialists in the diagnosis of Oral Potentially Malignant Disorders and Oral Cancer using Store-and-Forward technology.

Oral Cancer is one of the most common causes of morbidity and mortality. Screening and mobile Health (mHealth) based approach facilitates remote early detection of Oral cancer in a resource-constrained settings. The emerging eHealth technology has aided specialist reach to rural areas enabling remote monitoring and triaging to downstage Oral cancer. Though the diagnostic accuracy of the remote specialist has been evaluated, there are no studies evaluating the consistency among the remote specialists, to the best of our knowledge. The purpose of the study was to evaluate the interobserver agreement between the specialists through telemedicine systems in real-world settings using store and forward technology. Two remote specialists independently diagnosed the clinical images from image repositories, and the diagnostic accuracy was compared with onsite specialist and histopathological diagnosis when available. Moderate agreement (k = 0.682) between two remote specialists and (k = 0.629) between the onsite specialist and two remote specialists in diagnosing oral lesions. The sensitivity and specificity of remote specialist 1 were 92.7% and 83.3%, whereas remote specialist 2 was 95.8% and 60%, respectively, compared to histopathology. The store and forward technology and telecare can be effective tools in triaging and surveillance of patients.

Prevalence and determinants of hypertension among urban tribal communities in Nagaland, India-A community-based study.

Background: There is a paucity of data on the burden and factors associated with hypertension among the Nagas (collective term for tribal ethnic groups predominantly residing in Nagaland) living in an urban environment. Insights from this study will aid in mapping focused community-based and primary care interventions for hypertension. Objectives: To determine the prevalence and risk factors associated with hypertension among Nagas aged 30-50 years residing in urban Dimapur, Nagaland. Methods: A community-based cross-sectional study was conducted between January and July 2019. This study screened 660 participants for hypertension using a digital blood pressure apparatus. A semi-structured questionnaire was used to assess the risk factors, and anthropometric measurements were recorded using standard guidelines. Results: The prevalence of hypertension and pre-hypertension was 25.9% and 44.5%, respectively. Non-modifiable risk factors such as male gender (adjusted odds ratio [AOR]: 2.02; 95% confidence interval [CI]: 1.32-3.09), age > 40 years (AOR: 2.32; 95% CI: 1.57-3.41), family history of hypertension (AOR, 1.87, 95% CI: 1.19-2.92) and modifiable risk factors such as current alcohol consumption (AOR: 2.05; 95% CI: 1.27-3.31), high/very high perceived stress (AOR: 2.15; 95% CI: 1.28-3.62), lack of participation in stress relief activities (AOR: 2.08; 95% CI: 1.17-3.71) and overweight/obesity (AOR: 2.26; 95% CI: 1.55-3.30) were independently associated with hypertension in this study. Conclusion: To avert an impending health crisis in this community, a multipronged approach involving primary-care/family physicians, culturally appropriate awareness, and targeted community-based screening programs with an adept referral system must be implemented to curtail this emerging threat.

Overestimation of school-based deworming coverage resulting from school-based reporting.

BACKGROUND: Soil Transmitted Helminths (STH) infect over 1.5 billion people globally and are associated with anemia and stunting, resulting in an annual toll of 1.9 million Disability-Adjusted Life Years (DALYs). School-based deworming (SBD), via mass drug administration (MDA) campaigns with albendazole or mebendazole, has been recommended by the World Health Organization to reduce levels of morbidity due to STH in endemic areas. DeWorm3 is a cluster-randomized trial, conducted in three study sites in Benin, India, and Malawi, designed to assess the feasibility of interrupting STH transmission with community-wide MDA as a potential strategy to replace SBD. This analysis examines data from the DeWorm3 trial to quantify discrepancies between school-level reporting of SBD and gold standard individual-level survey reporting of SBD. METHODOLOGY/PRINCIPAL FINDINGS: Population-weighted averages of school-level SBD calculated at the cluster level were compared to aggregated individual-level SBD estimates to produce a Mean Squared Error (MSE) estimate for each study site. In order to estimate individual-level SBD coverage, these MSE values were applied to SBD estimates from the control arm of the DeWorm3 trial, where only school-level reporting of SBD coverage had been collected. In each study site, SBD coverage in the school-level datasets was substantially higher than that obtained from individual-level datasets, indicating possible overestimation of school-level SBD coverage. When applying observed MSE to project expected coverages in the control arm, SBD coverage dropped from 89.1% to 70.5% (p-value < 0.001) in Benin, from 97.7% to 84.5% (p-value < 0.001) in India, and from 41.5% to 37.5% (p-value < 0.001) in Malawi. CONCLUSIONS/SIGNIFICANCE: These estimates indicate that school-level SBD reporting is likely to significantly overestimate program coverage. These findings suggest that current SBD coverage estimates derived from school-based program data may substantially overestimate true pediatric deworming coverage within targeted communities. TRIAL REGISTRATION: NCT03014167.

Interpretable and Reliable Oral Cancer Classifier with Attention Mechanism and Expert Knowledge Embedding via Attention Map.

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.

Impact of adverse events during community-wide mass drug administration for soil-transmitted helminths on subsequent participation-a Theory of Planned Behaviour analysis.

BACKGROUND: Experiencing adverse events (AEs) during mass drug administration (MDA) could affect participation in future MDAs. This study aims to understand the potential influence of AEs during a community-wide MDA (cMDA) trial for soil-transmitted helminths (STH) in India on intention to participate in future cMDAs. METHODS: This study was conducted using a multi-method quantitative and qualitative approach among 74 participants who experienced an AE during STH cMDA and the 12 participants who subsequently refused cMDA treatment of the ongoing DeWorm3 trial. Path analysis and thematic analysis guided by the Theory of Planned Behaviour, was used. PRINCIPAL FINDINGS: Among 74 individuals who reported an AE, 12% refused treatment in the cMDA immediately subsequent to their AE and 4% refused in all subsequent cMDAs. Of these 74 individuals, 59 (80%) completed a survey and eight participated in in-depth interviews. A positive attitude towards deworming and perceived ability to participate in cMDA (perceived behavioural control) were significant predictors of intention to participate in cMDA (p<0.05). A positive attitude towards cMDA was associated with caste (χ2 = 3.83, P = 0.05), particularly among the scheduled caste/scheduled tribe (SC/ST) (62%). Perceived behavioural control in cMDA participation was associated with occupation (χ2 = 5.02, P<0.05), with higher perceived control among those engaged in skilled occupations (78%). Intention to participate in subsequent cMDAs was associated with caste and family type (χ2 = 3.83, P = 0.05 and χ2 = 7.50, P<0.05 respectively) and was higher among SC/ST (62%) and those with extended families (67%). In-depth interviews demonstrated that perceived severe AEs may lead to treatment refusal in future, particularly if children were affected. CONCLUSIONS: Intention to participate in future STH cMDAs was associated with caste (SC/ST) and family type (extended families). Therefore, community mobilization messages about potential AEs and their management may need to intentionally target non-SC/ST households, nuclear families, and those engaged in unskilled occupations to increase cMDA participation given the possibility of AEs occurring. TRIAL REGISTRATION: NCT03014167, ClinicalTrials.gov.

Exploring uncertainty measures in convolutional neural network for semantic segmentation of oral cancer images.

Significance: Oral 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. Aim: We 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. Approach: This 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. Results: The 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. Conclusions: Our 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.

Field validation of deep learning based Point-of-Care device for early detection of oral malignant and potentially malignant disorders.

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.

Visual association cortex links cues with conjunctions of reward and locomotor contexts.

Postrhinal cortex (POR) and neighboring lateral visual association areas are necessary for identifying objects and interpreting them in specific contexts, but how POR neurons encode the same object across contexts remains unclear. Here, we imaged excitatory neurons in mouse POR across tens of days prior to and throughout initial cue-reward learning and reversal learning. We assessed responses to the same cue when it was rewarded or unrewarded, during both locomotor and stationary contexts. Surprisingly, a large class of POR neurons were minimally cue-driven prior to learning. After learning, distinct clusters within this class responded selectively to a given cue when presented in a specific conjunction of reward and locomotion contexts. In addition, another class contained clusters of neurons whose cue responses were more transient, insensitive to reward learning, and adapted over thousands of presentations. These two classes of POR neurons may support context-dependent interpretation and context-independent identification of sensory cues.

Impact of the COVID-19 National Lockdown on a Rural and Tribal Population of Tamil Nadu, Southern India: A Mixed-Methods Survey.

We assessed the impact of the national lockdown on a rural and tribal population in Tamil Nadu, southern India. A mixed-methods approach with a pilot-tested, semi-structured questionnaire and focus group discussions were used. The impact of the lockdown on health, finances, and livelihood was studied using descriptive statistics. Multivariable logistic regression was carried out to identify factors associated with households that borrowed loans or sold assets during the lockdown, and unemployment during the lockdown. Of the 607 rural and tribal households surveyed, households from comparatively higher socioeconomic quintiles (adjusted odds ratio [aOR], 1.84; 95% CI, 1.01-3.34), with no financial savings (aOR, 2.91; 95% CI, 1.17-7.22), and with larger families (aOR, 1.76; 95% CI, 1.22-2.53), took loans or sold assets during the lockdown. Previously employed individuals from rural households (aOR, 5.07; 95% CI, 3.30-7.78), lower socioeconomic households (aOR, 3.08; 95% CI, 1.74, 5.45), and households with no savings (aOR, 1.78; 95% CI, 1.30-2.44) became predominantly unemployed during the lockdown. Existing government schemes for the elderly, differently abled, and widows were shown to be accessible to 89% of the individuals requiring these schemes in our survey. During the focus group discussions, the limited reach of online classes for schoolchildren was noted and attributed to the lack of smartphones and poor Internet connectivity. Although the sudden, unannounced national lockdown was imposed to flatten the COVID-19 curve, aspects related to livelihood and financial security were affected for both the rural and tribal populations.

Interpretable deep learning approach for oral cancer classification using guided attention inference network.

SIGNIFICANCE: Convolutional 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. AIM: Develop 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. APPROACH: We 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. RESULTS: The 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. CONCLUSIONS: We demonstrate the effectiveness of our approach for more interpretable and reliable oral potentially malignant lesion and malignant lesion classification.

Risk Factors and Hazards in the Household Environment for Elevated Blood Lead Levels in Urban Preschool Children of Vellore: A Case-Control Approach in the MAL-ED Birth Cohort.

OBJECTIVES: To study the household environmental risk factors and hazards associated with elevated blood lead levels (EBLLs) in preschool children in an urban setting of Vellore, South India. METHODS: A case-control study within the MAL-ED (Etiology, Risk Factors and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development) birth cohort was conducted between January 2014 and January 2015. The study included 153 pre-school children: 87 cases and 66 controls with elevated and normal BLLs, respectively. A structured questionnaire was used to assess the sociodemographic profile, household environment, breastfeeding practices, children's habits, and the use of cosmetics in them. Household environmental samples of wall and door paint, floor dust, drinking water, and cosmetics were estimated for lead levels using gas flame atomic absorption spectrometry (FAAS). RESULTS: Children born with low birth weight, those living in houses painted at least once in the last five years and those residing in houses older than ten years had a higher odds of EBLLs [OR (95% CI) = 3.79 (1.24-11.1); 4.84 (1.42-16.53); 5.07 (2.06-12.46), and 2.58 (0.99-6.69)], respectively. Drinking water samples from both cases (88%) and controls (95%) had lead levels more than the Environmental Protection Agency (EPA), USA recommendation of 0.015 ppm. CONCLUSIONS: Low birth weight and the household environment pose important risk factors/hazards for elevated blood lead levels in urban preschool children. Multipronged interventions that include government legislations, household environmental modification, safe water supply, and community education are pivotal in reducing lead exposure in young children.

Prevalence and determinants of oral potentially malignant lesions using mobile health in a rural block, northeast India.

In a rural block in North East India, community health workers (CHW) empowered with a mobile phone-based application screened a total of 2,686 participants for Oral Potentially Malignant Lesions (OPMLs), and an oral medicine specialist recommended treatment remotely. Independent risk factors were determined using independent multiple logistic regression models. Nearly 700 (26%) participants were identified with OPMLs. The sensitivity, specificity, positive predictive values, negative predictive values and accuracy of the CHW was 70.3, 88.4, 66.8, 89.9% and 83.7% respectively. Male gender, married status, smokeless tobacco, paan, areca-nut and alcohol consumption were independent predictors of OPMLs, the burden of which in North East India can be attributed to the high consumption of tobacco and non-tobacco products. Such programmes, with the recommendations from remote specialists, will facilitate early detection in remote settings.

Bayesian deep learning for reliable oral cancer image 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.

Classification of imbalanced oral cancer image data from high-risk population.

SIGNIFICANCE: Early 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. AIM: To reduce the class bias caused by data imbalance. APPROACH: We 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. RESULTS: By 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. CONCLUSIONS: Experimental 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.

Mobile-based oral cancer classification for point-of-care screening.

SIGNIFICANCE: Oral 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. AIM: To develop a mobile-based dual-mode image classification method and customized Android application for point-of-care oral cancer detection. APPROACH: The 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. RESULTS: We 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. CONCLUSIONS: Our study demonstrates the effectiveness of a mobile-based approach for oral cancer screening in low-resource settings.