Integrated Tuberculosis and COVID-19 Activities in Karachi and Tuberculosis Case Notifications

As the COVID-19 pandemic surged, lockdowns led to the cancellation of essential health services. As part of our Zero TB activities in Karachi, we adapted our approach to integrate activities for TB and COVID-19 to decrease the impact on diagnosis and linkage to care for TB treatment. We implemented the following: (1) integrated COVID-19 screening and testing within existing TB program activities, along with the use of an artificial intelligence (AI) software reader on digital chest X-rays; (2) home delivery of medication; (3) use of telehealth and mental health counseling; (4) provision of PPE; (5) burnout monitoring of health workers; and (6) patient safety and disinfectant protocol. We used programmatic data for six districts of Karachi from January 2018 to March 2021 to explore the time trends in case notifications, the impact of the COVID-19 pandemic, and service adaptations in the city. The case notifications in all six districts in Karachi were over 80% of the trend-adjusted expected notifications with three districts having over 90% of the expected case notifications. Overall, Karachi reached 90% of the expected case notifications during the COVID-19 pandemic. The collaborative efforts by the provincial TB program and private sector partners facilitated this reduced loss in case notifications.

Enhancement of digital chest images using a modified Sobel edge detection algorithm

Computed tomography (CT) images are an essential factor in the diagnosing procedure for various diseases affecting the internal organs. Edge detection can be used for the appropriate enhancement of the lung CT scan images for the diagnosis of the various interstitial lung diseases (ILD). In order to solve the issues of edge detection provided by the traditional Sobel operator, the paper proposes a Sobel 12D edge detection algorithm which uses the additional direction templates for the better identification of the edge details. First, the vertical and horizontal directions available in the traditional Sobel operator are extended to few more directions (a total of 12 directions) which enhances the edge extraction ability. Next part, compute the edge detected image using the Sobel 12D, Laplace, Prewitt, Robert’s Cross and Scharr operators for edge detection separately. It is followed by image fusion method which optimizes the edge detection by combining the edge detected images obtained using the Sobel 12D approach and the Laplace operator. The experimental results shows that the proposed algorithms generates a better detection of the edges than the other edge detection operators.

Neonatal digital chest radiography– should we be using additional copper filtration?

Objectives: Copper filtration removes lower energy X-ray photons, which do not enhance image quality but would otherwise contribute to patient dose. This study explores the use of additional copper filtration for neonatal mobile chest imaging. Methods: A controlled factorial-designed experiment was used to determine the effect of independent variables on image quality and radiation dose. These variables included: copper filtration (0Cu, 0.1Cu and 0.2Cu), exposure factors, SID and image receptor position (direct +tray). Image quality was evaluated using absolute visual grading analysis (VGA) and contrast-to-noise ratio (CNR) and entrance surface dose (ESD) was derived using an ionising chamber within the central X-ray beam. Results: VGA, CNR and ESD significantly reduced (p < 0.01) when using added copper filtration. For 0.1Cu, the percentage reduction was much greater for ESD (60%) than for VGA (14%) and CNR (20%), respectively. When compared to the optimal combinations of parameters for incubator imaging using no copper filtration, an increase in kV and mAs when using 0.1mmCu resulted in better image quality at the same radiation dose (direct) or, equal image quality at reduced dose (in-tray). The use of 0.1mmCu for neonatal chest imaging with a corresponding increase in kV and mAs is therefore recommended. Conclusions: Using additional copper filtration significantly reduces radiation dose (at increased mAs) without a detrimental effect on image quality. Advances in knowledge: This is the first study, using an anthropomorphic phantom, to explore the use of additional Cu for DR neonatal chest imaging and therefore helps inform practice to standardise and optimise this imaging examination.

Gender-based differences in community-wide screening for pulmonary tuberculosis in Karachi, Pakistan: an observational study of 311 732 individuals undergoing screening

We describe gender-based differences in a community-wide TB screening programme in Karachi, Pakistan, in which 311 732 individuals were screened in mobile camps using symptom questionnaires and van-mounted digital chest X-ray, between 1 January 2018 and 31 December 2019. Only 22.4% (69 869) of camp attendees were women. Female attendees were less likely to have sputum collected and tested (31.5% (95% CI 30.4% to 32.7%) vs 38.5% (95% CI 37.6% to 39.1%)) or to initiate TB treatment (75.9% (95% CI 68.1% to 82.6%) vs 82.8% (95% CI 78.9% to 86.2%)), when indicated. Among the participants, the age-standardised prevalence of active TB was higher among women (prevalence ratio 1.4, 95% CI 1.1 to 1.7). These findings underscore the importance of integrating gender into the design and monitoring of TB screening programmes to ensure that women and men benefit equally from this important intervention.

Digital chest radiography enhances screening efficiency for pulmonary tuberculosis in primary health clinics, South Africa

Background Optimized tuberculosis (TB) screening in high burden settings is essential for case finding. We evaluated digital chest x-ray with computer-aided detection (CAD) software (d-CXR) for identifying undiagnosed TB in three primary health clinics in South Africa. Methods The cross-sectional study consented adults who were sequentially screened for TB using the World Health Organization (WHO) four symptom questionnaire and d-CXR. Participants reporting ≥1 TB symptom and/or CAD score ≥60 (suggestive of TB) provided two spot sputum for Xpert MTB/RIF Ultra (Xpert Ultra) and liquid culture testing respectively. TB yield (proportion of screened tested positive) and number needed to test [NNT] (no of tests to identify one TB patient) were calculated. Risk factors for microbiologically confirmed or presumed (on radiological grounds) were determined. Results Among 3041 participants, 45% (1356/3,041) screened positive on either d-CXR or symptoms. TB yield was 2.3% (71/3041) using Xpert Ultra and 2.7% (82/3041) using Xpert Ultra plus culture. Modelled TB yield (identified by Xpert Ultra) by screening approach was: 1.9% (59/3041) for d-CXR alone, 2.0% (62/3041) for symptoms alone and 2.3% (71/3041) for both. The NNT was 9.7 for d-CXR, 17.8 for symptoms and 19.1 for d-CXR and/or symptom. Males, those with previous TB, untreated HIV or unknown HIV status, and acute illness were at higher risk of developing TB. Conclusion d-CXR screening identified a similar yield of undiagnosed TB compared to symptom-based screening, however required fewer diagnostic tests. Due to its objective nature, d-CXR screening may improve case detection in clinics.

Prediction and Detection of COVID-19 from Chest X-Rays using Transfer Learning based Deep Convolutional Neural Networks

With the ongoing outbreak of the COVID-19 global pandemic, the research community still struggles to develop early and reliable prediction and detection mechanisms for this infectious disease. The commonly used RT-PCR test is not readily available in areas with limited testing facilities, and it lags in performance and timeliness. This paper proposes a deep transfer learning-based approach to predict and detect COVID-19 from digital chest radiographs. In this study, three pre-trained convolutional neural network-based models (VGG16, ResNet18, and DenseNet121) have been fine tuned to detect COVID-19 infected patients from chest X-rays (CXRs). The most efficient model is further used to identify the affected regions using an unsupervised gradient-based localization technique. The proposed system uses a classification approach (normal vs. COVID-19 vs. pneumonia vs. lung opacity) using three supervised classification algorithms followed by gradient-based localization. The training, validation and testing of the system are performed using 21165 CXR images (10192 normal, 1345 pneumonia, 3616 COVID-19, and 6012 lung opacity). Simulation and evaluation results are presented using standard performance metrics, viz, accuracy, sensitivity, and specificity.