Government of Karnataka

The automation of road damage detection and classification is vital for enhancing road safety and enabling cost-effective, proactive infrastructure maintenance. This study introduces an advanced transfer learning-based framework designed to achieve precise and efficient road damage classification, validated on the RDD 2022 dataset, which spans six geographically diverse regions—India, Japan, Czech Republic, Norway, the U.S, and China. To ensure both specialization and generalization, the framework employs a two-phase fine-tuning strategy. A region-specific model is trained on country-wise datasets to capture localized damage patterns with high precision, while a generalized model leverages a cross-regional dataset encompassing diverse damage classes from all six countries, enhancing adaptability and ensuring robust performance across varying road conditions. To extract high-dimensional, discriminative representations of road damage images, the framework integrates twelve state-of-the-art deep learning architectures, including VGG16, VGG19, Xception, ResNet50, ResNet101, ResNet152, InceptionV3, DenseNet121, EfficientNetB0, and transformer-based models such as Vision Transformer and Swin Transformer. These extracted features undergo further refinement through K-means clustering, which enhances classification accuracy by structuring the feature space. Additionally, LIME-based feature selection identifies the 100 most salient features, ensuring a balance between computational efficiency and classification performance. The optimized feature vectors serve as inputs for multiple Machine learning classifiers, including SVM, KNN, Decision Trees, Gaussian Naïve Bayes, Logistic Regression, MLP, AdaBoost, XGBoost, Random Forest, LGBM (Light Gradient Boosting Machine), and Extra Trees, ensuring scalable and robust classification across diverse datasets. By integrating deep learning-based feature extraction, clustering-based refinement, and interpretable feature selection, the proposed methodology establishes a new benchmark in automated road maintenance. This comprehensive approach offers a scalable, high-precision solution that effectively balances accuracy, efficiency, and adaptability, making it well-suited for both localized and generalized datasets.

Introduction: Ballistocardiography (BCG) is a technique of contactless monitoring of the body’s vital parameters. During ventricular systole, aortic blood flow produces minuscule body displacements which can be picked up by piezoelectric sensors placed under the patient’s mattress. The BCG monitor connected to the sensors analyses these body displacements and generates basic vital parameters, which are displayed on a mobile device or computer. BCG devices may demonstrate improved compliance due to its unobtrusive nature, as well as affordability compared to standard bedside monitors, although their reliability remains unclear. Aim: To evaluate the accuracy of patients’ vital parameters as measured by the BCG device (Dozee), by comparison with simultaneous readings on a reference device, the Philips IntelliVue MP20 bedside monitor. Materials and Methods: This two-month prospective, observational study was conducted at the Intensive Care Unit (ICU) of Manipal Hospital, Yeshwantpur, Bengaluru, Karnataka, India from May 2023 to June 2023. All patients admitted to the ICU were assessed for the study, excluding those who were haemodynamically unstable or agitated, as well as those receiving haemodialysis, resulting in 121 patients who met the inclusion criteria. Routine vital parameters were obtained from each device, including Heart Rate (HR), Respiratory Rate (RR), Systolic Blood Pressure (SBP), Diastolic Blood Pressure (DBP), Mean Arterial Pressure (MAP), Peripheral Oxygen Saturation (SpO2) and Temperature (Temp). Mean values from the Dozee device were similar to that of the bedside monitor using One Way Analysis of Variance (ANOVA) using F-test and a p-value<0.05 was taken as significant. Patients’ vitals were recorded at four-hour intervals during their stay in ICU. Results: A total of 121 patients were included with a mean age of 62.6±16.6 years. The readings of each parameter were obtained at four-hour intervals. About 900 - 1000 readings were obtained. There was a good agreement between the values measured by the BCG device and the bedside monitor, with a p-value of <0.001 across all parameters: HR (mean 86.9 vs 82.9 beats/min), RR (20.8 vs 19.6 cycles/min), SBP (124.8 vs 122.8 mmHg), DBP (68.0 vs 67.7 mmHg), MAP (87.0 vs 85.3 mmHg), SpO2 (96.2 vs 95.1 %) and Temp (97.5 vs 96.9 ° F). However, reliability of the data varied across the parameters, with the most accuracy noted only in HR and BP readings. The BCG device also failed to give readings in about 44 (4.4%) of 1000 readings obtained on the bedside monitor. Conclusion: BCG devices can substitute for regular bedside monitors, although it is reliable only for measuring three vital parameters, HR, BP and RR. Although the contactless technology allows for better patient compliance, the present study suggests that it is safe for use only in bedbound patients who are relatively stable, or in conditions where bedside monitors or nursing care are unfeasible or unavailable.

Background:People with tuberculosis (TB) have a suicide risk up to 25 times higher than the general population. It may be due to stigma, isolation, prolonged treatment, and psychological distress from the disease. Hence, empowering healthcare workers (HCWs) who are involved in TB care with suicide prevention skills is essential. Novelty:This study will be integrating suicide prevention into TB care and will be evaluating an asynchronous learning management solution (LMS)-based training model against traditional in-person gatekeeper training for HCWs of high-risk population. Objectives:This study will examine the effectiveness of the gatekeeper suicide prevention training online e-module compared to in-person training for National Tuberculosis Elimination Program (NTEP) HCWs. Methods:HCWs in the Karnataka State NTEP program will be randomly assigned to two training groups using parallel cluster-randomization at the district level. The online training group (intervention group) will receive training through customized and locally adapted e-modules hosted on an LMS. The In-person training group (control group) will undergo training through presentations, role-play simulations, and discussions. Training effectiveness will be assessed using the Kirkpatrick model, with statistical evaluation and comparison of pre- and post-test scores to measure knowledge acquisition, skill retention, and application (number of referrals) across both groups. Expected Outcome:The study anticipates that e-module training will be as effective as in-person training for developing competence and confidence in suicide prevention and early identification and referral of suicidality.

Abstract Background Neurological disorders are leading causes of disability and death, with disproportionate burden in low- and middle-income countries. The World Health Organization’s Intersectoral Global Action Plan on Epilepsy and Other Neurological Disorders prioritises awareness, risk reduction and strengthened care pathways within a life-course brain health agenda. Objective To describe the design, implementation strategies and outputs of the Karnataka Brain Health Initiative (KaBHI) pilot’s brain health promotion and awareness component. Methods This TIDieR-informed descriptive implementation report (Perspective) summarises activities delivered January 2022–May 2023 in three districts of Karnataka, India. Programme documents and monitoring records were synthesised for rationale, audiences, materials, delivery channels, settings, implementers and activity-level outputs, interpreted using selected RE-AIM domains. Results KaBHI developed Kannada- and English-language information, education and communication materials addressing selected neurological conditions and brain health messages. Delivery used radio and YouTube dissemination, public ambassador engagement, theme-based awareness days, community screening camps, participatory DrumJam sessions, school and workplace outreach, and frontline health-worker sensitisation. Outputs included ~ 800 World Brain Day participants (250 dementia-screenings); 467 screened on World Stroke Day (101 requiring counselling/referral; 63 referred, including 27 aged < 30 years); > 150 World Alzheimer’s Day participants; 12 DrumJam sessions (~ 200 participants); and > 1,000 community camp attendees. Indirect reach via broadcast and online posting was documented; deduplicated reach and longitudinal outcomes were not consistently available. Conclusion KaBHI demonstrates a feasible pilot-scale multisectoral approach to embedding brain health promotion within a state neurological public health initiative, supporting future evaluation of reach, equity, stigma, care-seeking, referral completion, rehabilitation uptake and longer-term outcomes.

Background and objectives Accredited social health activists (ASHAs) play a key role in community mental health services in Karnataka; however, evidence on training interventions has not been systematically reviewed. We aimed to systematically synthesise evidence on mental health training interventions for ASHAs in Karnataka, focusing on provider competencies and service delivery outcomes. Methods Systematic search of databases, trial registries, and grey literature (October 2025-January 2026) included studies published between 2000 and 2025. Eligible studies involved ASHAs in Karnataka and evaluated mental health training interventions with outcomes related to provider competencies or service delivery. Due to heterogeneity, findings were synthesised narratively using SWiM guidelines for quantitative studies and GRADE-CERQual for qualitative. The risk of bias was assessed using RoB 2, ROBINS-I (v2), MMAT, and CASP. Results Of 648 records screened, 13 studies comprising 567 ASHAs met the inclusion criteria. Narrative synthesis showed improved provider competencies and service delivery outcomes from quantitative studies, while qualitative studies revealed moderate confidence. Overall, risk of bias was judged as 'low to some concerns'. Digital and hybrid models showed more sustained gains than standalone in-person training. Selected service delivery outcomes, including screening coverage, home visits and supervisory engagement, improved with structured support. However, the evidence was limited by pre-post designs, small sample size, reliance on self-reported outcomes, and heterogeneity in interventions and outcome measures. Interpretation and conclusion Structured training may improve ASHAs' mental health provider competencies and service delivery. Digital and hybrid models are promising, but evidence remains limited, highlighting the need for rigorous studies with standardised outcomes and long-term follow up.