Building Radiology Equity: Themes from the 2023 RAD-AID Conference on International Radiology and Global Health.

Low- and middle-income countries are significantly impacted by the global scarcity of medical imaging services. Medical imaging is an essential component for diagnosis and guided treatment, which is needed to meet the current challenges of increasing chronic diseases and preparedness for acute-care response. We present some key themes essential for improving global health equity, which were discussed at the 2023 RAD-AID Conference on International Radiology and Global Health. They include (1) capacity building, (2) artificial intelligence, (3) community-based patient navigation, (4) organizational design for multidisciplinary global health strategy, (5) implementation science, and (6) innovation. Although not exhaustive, these themes should be considered influential as we guide and expand global health radiology programs in low- and middle-income countries in the coming years.

MRI scarcity in low- and middle-income countries.

Since the introduction of MRI as a sustainable diagnostic modality, global accessibility to its services has revealed a wide discrepancy between populations-leaving most of the population in LMICs without access to this important imaging modality. Several factors lead to the scarcity of MRI in LMICs; for example, inadequate infrastructure and the absence of a dedicated workforce are key factors in the scarcity observed. RAD-AID has contributed to the advancement of radiology globally by collaborating with our partners to make radiology more accessible for medically underserved communities. However, progress is slow and further investment is needed to ensure improved global access to MRI.

Deploying Artificial Intelligence for Thoracic Imaging Around the World.

PURPOSE: Artificial intelligence (AI) thoracic imaging applications are increasingly being deployed in low- and middle-income countries (LMICs). Radiologists have a critical gatekeeping role to ensure the effective and ethical implementation of AI solutions. RAD-AID International uses a three-pronged implementation strategy to overcome challenges pervasive in LMICs. METHODS: During a similar period, an AI software for chest radiography (CXR) interpretation was deployed at two tertiary hospitals located in Guyana and Nigeria. The three-pronged implementation strategy of clinical education, infrastructure implementation, and phased AI introduction was used. A PACS with a cloud component was installed at each institution. Radiology residents and attending physicians at these institutions completed an introduction-to-AI course to prime them for the use of AI solutions. A phased introduction of the AI software was performed to allow local validation as well as trust building and workflow integration. Local validation processes were used at each site by comparing AI outputs with standardized prospectively generated reports by local radiologists and study team members, allowing for slight differences in the goals of AI software use between sites. RESULTS: The PACS was successfully installed at both institutions. Thirty participants completed the introduction-to-AI course with an average pre-knowledge test score of 75% and an average posttest score of 95%. The focus of the validation process at various sites was reflective of the intended use of the AI software. In Guyana, it revealed an 87% concordance rate between radiologists and the AI model for triaging normal versus abnormal findings on CXR. In Nigeria, an 85% concordance rate between radiologists and the AI model for reporting tuberculosis on CXR was noted. The AI software was successfully deployed and is being used as intended at both institutions. CONCLUSIONS: There are unique barriers to the adoption of AI in LMICs requiring an implementation strategy in collaboration with local institutions and industry partners to ensure success.

Estimating Catchment Populations of Global Health Radiology Outreach Using Geographic Information Systems Analysis.

PURPOSE: The purpose of this study was to design, develop, and test geographic information systems (GIS) analytic methods for quantifying and characterizing catchment populations across all sites served by a radiology global health organization. METHODS: The analysis included populations served by 78 low-resource medical facilities in 32 countries partnered with radiology nonprofit organization, RAD-AID International. Three constraints were used to approximate patient catchment areas: (1) 1-hour driving time, (2) 1-hour walking time, and (3) 10-mile circular radius. GIS calculated populations within each constraint using publicly available geospatial input databases, including a global digital elevation model, population and land cover data, and road locations from OpenStreetMap. Demographic and health data from the World Health Organization were incorporated to provide further characteristics of covered populations. RESULTS: The total populations served by all RAD-AID sites as measured by driving time, walking time, and 10-mile radius were 189,241,193 (47.8% female), 26,190,117 (48.7% female), and 110,884,095 (48.1% female), respectively. For individual locations, median population within 1-hour driving time was 1,795,977 (range: 8,742-30,630,800), with an average life expectancy of 68.4 ± 5.8 years. Median child mortality before age 5 was 3.8% (range: 0.9%-8.3%), and median prevalence of human immunodeficiency virus infection was 3.1% (range: 0.7%-10.9%). CONCLUSION: In this study, GIS provided a robust multisite analysis for estimating the potential global population reached by an international radiology outreach organization with targeted individual site measurements. Given heightened needs to accurately characterize global outreach populations, this GIS-based approach may be useful for analysis, outreach planning, and resource allocation among global health organizations.

Strategic Radiology Outreach Planning for Underserved Populations Using Geographic Information Systems.

PURPOSE: Geographic information systems (GIS) are widely used in public health research but rarely used in radiology research. GIS can be an impactful tool in radiology global health to locate medically underserved populations and poor transportation infrastructure, characterize medical needs, and design outreach programs. Using the example of aircraft-based outreach in Alaska, we demonstrate the utility of GIS in radiological program planning for global health. METHODS: Multicriteria GIS evaluations were performed to create a health severity index, using life expectancy and percentage uninsured data, and an accessibility severity index, using distance from roads and health centers or hospitals. These indices were combined with population density to create a final health access severity index (HASI). A map presenting suitable hybrid airship operating areas was produced using land cover data. Alaskan health care facilities were georeferenced to create a coordinate data set. Infrastructure was obtained from OpenStreetMap. Health data were accessed from the 2017 American Community Survey and CDC US Small-area Life Expectancy Estimates Project. RESULTS: GIS analyzed 738,050 Alaskans. The health severity index identified decreased health outcomes (high or very high severity) in 285,446 (39%) Alaskans, and the accessibility severity index determined decreased access to care in 218,201 (30%). Combined, the HASI established 165,108 (22%) Alaskans as underserved with high or very high overall severity. Thirty-nine percent of Alaska land area is suitable for hybrid airship operations, including 27% of HASI high and very high severity areas. CONCLUSIONS: GIS identified underserved populations for mobile radiology outreach in Alaska and may be useful for global health outreach planning and resource allocation.

Artificial Intelligence in Low- and Middle-Income Countries: Innovating Global Health Radiology.

Scarce or absent radiology resources impede adoption of artificial intelligence (AI) for medical imaging by resource-poor health institutions. They face limitations in local equipment, personnel expertise, infrastructure, data-rights frameworks, and public policies. The trustworthiness of AI for medical decision making in global health and low-resource settings is hampered by insufficient data diversity, nontransparent AI algorithms, and resource-poor health institutions' limited participation in AI production and validation. RAD-AID's three-pronged integrated strategy for AI adoption in resource-poor health institutions is presented, which includes clinical radiology education, infrastructure implementation, and phased AI introduction. This strategy derives from RAD-AID's more-than-a-decade experience as a nonprofit organization developing radiology in resource-poor health institutions, both in the United States and in low- and middle-income countries. The three components synergistically provide the foundation to address health care disparities. Local radiology personnel expertise is augmented through comprehensive education. Software, hardware, and radiologic and networking infrastructure enables radiology workflows incorporating AI. These educational and infrastructure developments occur while RAD-AID delivers phased introduction, testing, and scaling of AI via global health collaborations.

Overcoming Challenges for Successful PACS Installation in Low-Resource Regions: Our Experience in Nigeria.

In this paper, we walk you through our challenges, successes, and experience while participating in a Global Health Outreach Project at the University College Hospital (UCH) Ibadan, Nigeria. The scope of the project was to install a Picture Archive and Communication System (PACS) to establish a centralized viewing network at UCH's Radiology Department, for each of their digital modalities. Installing a PACS requires robust servers, the ability to retrieve and archive studies, ensuring workstations can view studies, and the configuration of imaging modalities to send studies. We anticipated that we might experience hurdles for each of these requirements, due to limited resources and without the availability to make a site visit prior to the start of the project. While we ultimately experienced delays and troubleshooting was required at each turn of the install, with the help of dedicated volunteers both on and off-site and the UCH staff, our shared goal was accomplished.

A Global Perspective on Screening.

Breast cancer is emerging as a major global public health problem. Incidence and mortality continues to rise in low- and middle-income countries (LMICs). A significant and growing disparity exists between high-income countries and LMICs in the availability of screening services and associated preventable mortality. However, population imaging-based screening programs are not appropriate for all settings. Planners should perform a thorough assessment of the target setting prior to implementing any breast cancer detection program, as appropriate guidelines vary according to the resources available. Financial, social, and cultural barriers to breast cancer care need to be addressed to sustainably improve the morbidity and mortality of the populations and make efficient use of available services. Creative approaches, such as mobile and portable imaging and bundling of services, can facilitate the installation of early breast cancer detection programs in LMICs. While image-based screening programs are not initially resource-appropriate in many LMICs, planners can work towards this goal as part of their comprehensive breast cancer detection strategy.

Ultrasound for Breast Cancer Detection Globally: A Systematic Review and Meta-Analysis.

PURPOSE: Mammography is not always available or feasible. The purpose of this systematic review and meta-analysis is to assess the diagnostic performance of ultrasound as a primary tool for early detection of breast cancer. MATERIALS AND METHODS: For this systematic review and meta-analysis, we comprehensively searched PubMed and SCOPUS to identify articles from January 2000 to December 2018 that included data on the performance of ultrasound for detection of breast cancer. Studies evaluating portable, handheld ultrasound as an independent detection modality for breast cancer were included. Quality assessment and bias analysis were performed with the Quality Assessment of Diagnostic Accuracy Studies-2 tool. Sensitivity analyses and meta-regression were used to explore heterogeneity. The study protocol has been registered with the international prospective register of systematic reviews (PROSPERO identifier: CRD42019127752). RESULTS: Of the 526 identified studies, 26 were eligible for inclusion. Ultrasound had an overall pooled sensitivity and specificity of 80.1% (95% CI, 72.2% to 86.3%) and 88.4% (95% CI, 79.8% to 93.6%), respectively. When only low- and middle-income country data were considered, ultrasound maintained a diagnostic sensitivity of 89.2% and specificity of 99.1%. Meta-analysis of the included studies revealed heterogeneity. The high sensitivity of ultrasound for the detection of breast cancer was not statistically significantly different in subgroup analyses on the basis of mean age, risk, symptoms, study design, bias level, and study setting. CONCLUSION: Given the increasing burden of breast cancer and infeasibility of mammography in certain settings, we believe these results support the potential use of ultrasound as an effective primary detection tool for breast cancer, which may be beneficial in low-resource settings where mammography is unavailable.

Lymphocyte-driven regional immunopathology in pneumonitis caused by impaired central immune tolerance.

Autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED), a monogenic disorder caused by AIRE mutations, presents with several autoimmune diseases. Among these, endocrine organ failure is widely recognized, but the prevalence, immunopathogenesis, and treatment of non-endocrine manifestations such as pneumonitis remain poorly characterized. We enrolled 50 patients with APECED in a prospective observational study and comprehensively examined their clinical and radiographic findings, performed pulmonary function tests, and analyzed immunological characteristics in blood, bronchoalveolar lavage fluid, and endobronchial and lung biopsies. Pneumonitis was found in >40% of our patients, presented early in life, was misdiagnosed despite chronic respiratory symptoms and accompanying radiographic and pulmonary function abnormalities, and caused hypoxemic respiratory failure and death. Autoantibodies against BPIFB1 and KCNRG and the homozygous c.967_979del13 AIRE mutation are associated with pneumonitis development. APECED pneumonitis features compartmentalized immunopathology, with accumulation of activated neutrophils in the airways and lymphocytic infiltration in intraepithelial, submucosal, peribronchiolar, and interstitial areas. Beyond APECED, we extend these observations to lung disease seen in other conditions with secondary AIRE deficiency (thymoma and RAG deficiency). Aire-deficient mice had similar compartmentalized cellular immune responses in the airways and lung tissue, which was ameliorated by deficiency of T and B lymphocytes. Accordingly, T and B lymphocyte-directed immunomodulation controlled symptoms and radiographic abnormalities and improved pulmonary function in patients with APECED pneumonitis. Collectively, our findings unveil lung autoimmunity as a common, early, and unrecognized manifestation of APECED and provide insights into the immunopathogenesis and treatment of pulmonary autoimmunity associated with impaired central immune tolerance.

Holistic classification of CT attenuation patterns for interstitial lung diseases via deep convolutional neural networks.

Interstitial lung diseases (ILD) involve several abnormal imaging patterns observed in computed tomography (CT) images. Accurate classification of these patterns plays a significant role in precise clinical decision making of the extent and nature of the diseases. Therefore, it is important for developing automated pulmonary computer-aided detection systems. Conventionally, this task relies on experts' manual identification of regions of interest (ROIs) as a prerequisite to diagnose potential diseases. This protocol is time consuming and inhibits fully automatic assessment. In this paper, we present a new method to classify ILD imaging patterns on CT images. The main difference is that the proposed algorithm uses the entire image as a holistic input. By circumventing the prerequisite of manual input ROIs, our problem set-up is significantly more difficult than previous work but can better address the clinical workflow. Qualitative and quantitative results using a publicly available ILD database demonstrate state-of-the-art classification accuracy under the patch-based classification and shows the potential of predicting the ILD type using holistic image.

Joint solution for PET image segmentation, denoising, and partial volume correction.

Segmentation, denoising, and partial volume correction (PVC) are three major processes in the quantification of uptake regions in post-reconstruction PET images. These problems are conventionally addressed by independent steps. In this study, we hypothesize that these three processes are dependent; therefore, jointly solving them can provide optimal support for quantification of the PET images. To achieve this, we utilize interactions among these processes when designing solutions for each challenge. We also demonstrate that segmentation can help in denoising and PVC by locally constraining the smoothness and correction criteria. For denoising, we adapt generalized Anscombe transformation to Gaussianize the multiplicative noise followed by a new adaptive smoothing algorithm called regional mean denoising. For PVC, we propose a volume consistency-based iterative voxel-based correction algorithm in which denoised and delineated PET images guide the correction process during each iteration precisely. For PET image segmentation, we use affinity propagation (AP)-based iterative clustering method that helps the integration of PVC and denoising algorithms into the delineation process. Qualitative and quantitative results, obtained from phantoms, clinical, and pre-clinical data, show that the proposed framework provides an improved and joint solution for segmentation, denoising, and partial volume correction.

Cough Frequency During Treatment Associated With Baseline Cavitary Volume and Proximity to the Airway in Pulmonary TB.

BACKGROUND: Cough frequency, and its duration, is a biomarker that can be used in low-resource settings without the need of laboratory culture and has been associated with transmission and treatment response. Radiologic characteristics associated with increased cough frequency may be important in understanding transmission. The relationship between cough frequency and cavitary lung disease has not been studied. METHODS: We analyzed data in 41 adults who were HIV negative and had culture-confirmed, drug-susceptible pulmonary TB throughout treatment. Cough recordings were based on the Cayetano Cough Monitor, and sputum samples were evaluated using microscopic observation drug susceptibility broth culture; among culture-positive samples, bacillary burden was assessed by means of time to positivity. CT scans were analyzed by a US-board-certified radiologist and a computer-automated algorithm. The algorithm evaluated cavity volume and cavitary proximity to the airway. CT scans were obtained within 1 month of treatment initiation. We compared small cavities (≤ 7 mL) and large cavities (> 7 mL) and cavities located closer to (≤ 10 mm) and farther from (> 10 mm) the airway to cough frequency and cough cessation until treatment day 60. RESULTS: Cough frequency during treatment was twofold higher in participants with large cavity volumes (rate ratio [RR], 1.98; P = .01) and cavities located closer to the airway (RR, 2.44; P = .001). Comparably, cough ceased three times faster in participants with smaller cavities (adjusted hazard ratio [HR], 2.89; P = .06) and those farther from the airway (adjusted HR, 3.61;, P = .02). Similar results were found for bacillary burden and culture conversion during treatment. CONCLUSIONS: Cough frequency during treatment is greater and lasts longer in patients with larger cavities, especially those closer to the airway.

Quantitative Image Quality Comparison of Reduced- and Standard-Dose Dual-Energy Multiphase Chest, Abdomen, and Pelvis CT.

We present a new image quality assessment method for determining whether reducing radiation dose impairs the image quality of computed tomography (CT) in qualitative and quantitative clinical analyses tasks. In this Institutional Review Board-exempt study, we conducted a review of 50 patients (male, 22; female, 28) who underwent reduced-dose CT scanning on the first follow-up after standard-dose multiphase CT scanning. Scans were for surveillance of von Hippel-Lindau disease (N = 26) and renal cell carcinoma (N = 10). We investigated density, morphometric, and structural differences between scans both at tissue (fat, bone) and organ levels (liver, heart, spleen, lung). To quantify structural variations caused by image quality differences, we propose using the following metrics: dice similarity coefficient, structural similarity index, Hausdorff distance, gradient magnitude similarity deviation, and weighted spectral distance. Pearson correlation coefficient and Welch 2-sample t test were used for quantitative comparisons of organ morphometry and to compare density distribution of tissue, respectively. For qualitative evaluation, 2-sided Kendall Tau test was used to assess agreement among readers. Both qualitative and quantitative evaluations were designed to examine significance of image differences for clinical tasks. Qualitative judgment served as an overall assessment, whereas detailed quantifications on structural consistency, intensity homogeneity, and texture similarity revealed more accurate and global difference estimations. Qualitative and quantitative results indicated no significant image quality degradation. Our study concludes that low(er)-dose CT scans can be routinely used because of no significant loss in quantitative image information compared with standard-dose CT scans.