Multicenter Chest Radiographic Findings from Ethiopia with Radiologist-Generated Annotations

Multicenter Chest Radiographic Findings from Ethiopia with Radiologist-Generated Annotations | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multicenter Chest Radiographic Findings from Ethiopia with Radiologist-Generated Annotations Messay Gebrekidan, Samuel S. Hailu, Michael Teklehaimanot Abera, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9417593/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background Chest radiography is widely used in Ethiopia for the evaluation of respiratory and cardiac disease. However, chest X-ray datasets used to develop and benchmark artificial intelligence (AI) systems are predominantly derived from high-income settings, which may limit generalizability and obscure region-specific radiographic patterns. Purpose To describe multicenter patterns of chest X-ray imaging findings in Ethiopia using the Afro-Chest X-ray cohort and to summarize the radiologist-led workflow used to generate high-quality localization labels. Materials and Methods This retrospective multicenter study included deidentified chest X-rays acquired at 10 Ethiopian institutions from December 2022 through July 2025. After quality filtering, 55,409 chest X-rays were retained. Radiology reports were standardized using the Afro-Chest X-ray reporting template (Table 5 ). A stratified subset of 11,880 chest X-rays was manually annotated by 11 radiologists using bounding boxes for 19 thoracic findings (Table 3) with recorded confidence levels. Finding patterns was summarized descriptively using counts and proportions at the exam and finding-instance levels. Results Among 55,409 chest X-rays, 31,939 were linked to radiology reports from 48,962 patients (male, 18,324 [37.4%]; female, 30,387 [62.1%]). In the annotated subset (11,880 chest X-rays), 7,003 (58.9%) were abnormal and contained 22,531 labeled finding instances (mean, 3.22 instances per abnormal chest X-ray). The most frequent findings were reticulonodular or ground-glass opacities (n = 6,896, 30.6%), pleural effusion (n = 2,685, 11.9%), cardiomegaly or chamber enlargement (n = 2,161, 9.6%), fibrosis or fibrobronchiectatic change (n = 1,869, 8.3%), and consolidation with cavitary lesions (n = 1,835, 8.1%). Conclusion In this multicenter Ethiopian cohort, annotated abnormalities were dominated by parenchymal opacities, pleural effusion, and cardiomegaly. Afro-Chest X-ray provides a radiologist-verified reference for describing regional chest X-ray patterns and supporting the local validation and development of AI systems applicable to East and other sub-Saharan African clinical settings. Cardiac & Cardiovascular Systems Artificial Intelligence and Machine Learning Tropical Medicine Chest X-ray Ethiopia multicenter cohort dataset artificial intelligence radiology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Advances in Knowledge This multicenter Ethiopian cohort demonstrates that reticulonodular or ground-glass opacities, pleural effusion, cardiomegaly, and fibrotic changes account for most chest X-ray abnormalities in routine clinical practice. The Afro-Chest X-ray provides a large, radiologist-annotated reference dataset from an under-represented East African population, enabling characterization of regional chest radiographic patterns and supporting improved generalizability of imaging research and AI systems beyond high-income settings. Implications for Patient Care Region-specific chest X-ray patterns underscore the need for local validation before deployment of AI interpretation tools to ensure accurate and safe use in patient care. A curated, radiologist-verified Ethiopian chest X-ray cohort provides a reference standard that can support local performance assessment, quality improvement, and adaptation of AI tools for high-burden cardiopulmonary conditions in East African healthcare settings. Introduction Chest radiography is the most accessible and widely used imaging modality in Ethiopia and is central to the evaluation of respiratory and cardiac disease. Globally, an estimated 3.6 billion diagnostic radiology examinations are performed annually, underscoring the central role of X-ray imaging in healthcare systems [14]. In parallel, AI methods for chest X-ray interpretation have advanced rapidly, largely driven by the availability of large, publicly accessible datasets primarily derived from high-income settings [1–5]. The under-representation of African clinical imaging data, characterized by distinct image acquisitions and region-specific disease patterns, may reduce the generalizability of AI models [9, 10]. In Ethiopia, AI-enabled radiology is emerging; however, publicly described, multicenter, chest X-ray cohorts curated by radiologists remain scarce [6]. The Afro-Chest X-ray initiative was developed to address this gap through the creation of a multicenter Ethiopian cohort with linked radiology reports and radiologist-generated localization labels for 19 findings (reference to table). The objectives of this study were to (a) describe national-level patterns of annotated chest X-ray findings in Ethiopia using the Afro-Chest X-ray cohort and (b) document the clinical workflow used to generate standardized, high-quality labels suitable for epidemiologic description and supervised AI development. This study focuses on imaging patterns and annotation methodology rather than disease prevalence or AI performance evaluation. Ethiopia, the second most populous nation in Africa, has approximately 600 radiology practitioners, including radiologists and radiology trainees, corresponding to roughly 0.92 radiologists per 200,000 population [13]. Similar workforce constraints are present across many African countries and contribute to high clinical workload and limited access to subspecialty expertise, with potential downstream effects on diagnostic quality, service, and radiologist burnout [10, 12]. These realities underscore the need for scalable, context-aware supportive technologies in diagnostic imaging. High-quality reference datasets that reflect local practice and disease patterns are essential to the development of such technology. Materials and Methods Study Design and Ethics This retrospective multicenter cohort study included deidentified chest X-rays obtained during routine clinical care at 10 Ethiopian institutions. National and institutional approvals were obtained through national and local ethics oversight pathways. Because the dataset was deidentified and collected retrospectively, informed consent was waived where applicable. Participating Institutions Participating institutions included a mix of public referral hospitals, teaching hospitals, and private diagnostic centers, primarily located in Addis Ababa, with additional data from a tertiary referral hospital in Mekelle (Tigray Region). This institutional composition was intended to reflect routine chest radiography practice across major Ethiopian imaging providers, while acknowledging that some geographic regions remain under-represented. Data Acquisition and Preprocessing Chest radiographs were exported from institution-specific PACS archives as DICOM files and underwent deidentification and quality filtering to remove non-chest images and studies unsuitable for analysis . Approximately 61,000 DICOM radiographs were retrieved, of which 55,409 chest X-rays were retained following preprocessing. Radiology Report Harmonization Reporting practices varied across participating institutions. To enable consistent label mapping and cross-site comparison, free-text reports were normalized to the Afro-Chest X-ray reporting template with standardized sections (Technique, Findings, Impression). Report terminology was mapped to a unified 19-label taxonomy through a predefined dictionary. In internal consistency checks, concordance between standardized report-derived labels and corresponding expert annotations exceeded 90% (90.4%). Expert Annotation Protocol A stratified subset of radiographs was selected for expert manual annotation to enrich the representation of abnormal findings. Eleven radiologists completed structured training and calibration on the VIA annotation tool [15] prior to labeling. Radiologists annotated 19 thoracic finding categories using bounding boxes and recorded confidence scores for each finding instance. Examinations were categorized as diagnostic or nondiagnostic to reflect image interpretability. The Radiologists were given the patient's gender and age for annotation, but no clinical history. To illustrate the structure of the manual labels, representative examples are provided of (a) a normal radiograph labeled “no findings” without bounding boxes and (b) an abnormal radiograph with bounding-box localization and an associated labeler confidence score (Figure 5). Representative uncommon entities in the cohort are shown in Figure 6, with panel-level diagnoses and key imaging findings summarized in Table S1. Quality Assurance A validation panel reviewed label definitions, adjudicated ambiguous cases, and enforced consistency of label usage across annotators. Discrepant examinations were resolved by consensus review. Statistical Analysis Analyses were descriptive. Categorical variables are reported as counts and percentages. In the annotated subset, finding frequencies are reported at the finding-instance level, allowing multiple findings per radiograph. Results Cohort Characteristics After preprocessing, the Afro-Chest X-ray cohort included 55,409 chest X-rays. Of these, 31,939 radiographs were linked to radiology reports from 48,971 unique patients (male, 18,324 [37.4%]; female, 30,387 [62.1%]; unspecified sex, 260 [0.5%]). Some patients have multiple examinations, reflecting follow-up imaging to see the response to treatments performed as part of routine clinical care. Annotated Subset and Abnormality Burden A total of 11,880 chest radiographs underwent radiologist annotation. Of these, 7,003 (58.9%) were classified as abnormal and contained 22,531 labeled finding instances (mean, 3.22 instances per abnormal radiograph). The remaining 4,877 examinations (41%) were classified as normal diagnostic examinations, and 180 radiographs were deemed non-diagnostic. National Patterns of Common Findings The five most frequent abnormality categories (Table 4) accounted for 15,446 of 22,531 labeled finding instances (68.6%). Reticulonodular or ground-glass opacity was the most frequently annotated finding. Pleural effusion and cardiomegaly were also common, and chronic fibrotic change and consolidation or cavitation comprised substantial portions of abnormal labels. Less frequently annotated findings included hernia, opaque hemithorax, mass, lung cyst, pneumothorax, and diaphragmatic or subdiaphragmatic abnormalities. Beyond common abnormalities, the Afro-Chest X-ray cohort contains uncommon and regionally important entities that are often underrepresented in widely used public chest X-ray datasets. Figure 6 illustrates representative examples, including pulmonary hydatid disease patterns (including a “water lily” configuration), diffuse metastatic nodules, a lung mass, esophageal dilation compatible with achalasia, and congenital lobar hyperinflation. These examples complement the national-level prevalence findings by highlighting the cohort’s phenotypic breadth and clinical diversity. Figure 1 summarizes cohort assembly. Figure 2 shows site distribution. Figure 3 shows the most frequent annotated findings. Figure 4 illustrates the Afro-Chest X-ray workflow. Figure 5Examples of Afro-Chest X-ray manual labeling. Figure 6 Unique and uncommon entities (multi-panel) Discussion In this multicenter Ethiopian cohort, annotated chest X-ray abnormalities were dominated by parenchymal opacities, pleural effusion, and cardiomegaly. This distribution is consistent with routine clinical imaging indications in Ethiopia, where infectious and inflammatory lung disease, pleural pathology, and cardiac conditions account for a substantial proportion of chest X-ray utilization [7, 8]. The prominence of fibrotic and fibrobronchiectatic changes further suggests that chronic or post-infectious lung disease is frequently encountered on chest X-rays interpreted across Ethiopian institutions. These radiographic findings have practical implications for both clinical training and AI development. First, they provide a data-driven characterization of the most common abnormality patterns encountered in participating facilities. Second, they highlight class-imbalance considerations relevant to AI development: while the most prevalent imaging patterns support robust model training, less common but clinically important findings may require targeted sampling strategies or collaborative data sharing to ensure adequate representation [11]. The Afro-Chest X-ray dataset was designed to support both classification and localization tasks by using radiologist-generated bounding box confidence scoring and a consensus-based quality assurance process [15]. To enable scalable annotation across institutions with heterogeneous reporting practices, free-text radiology reports were standardized to a shared reporting template. The labeling process was iterative and designed to support comprehensive abnormality detection. Annotations distinguished critical findings, such as active consolidation, from subcritical or chronic abnormalities. Label normalization incorporated finding-level prioritization, emphasizing clinically significant selected findings (e.g., pneumothorax, atelectasis, and pulmonary nodule). Limitations This study has limitations. Most chest radiographs were collected from institutions in Addis Ababa, with additional data from Mekelle, resulting in limited direct representation of other regions in Ethiopia. However, Addis Ababa serves as the country's largest national referral hub, receiving patients from across Ethiopia and neighboring East African countries (including Sudan, Eritrea, and Djibouti). Consequently, despite the geographic concentration of participating sites, the cohort likely captures a broad spectrum of national and regional disease patterns relevant to East African clinical practice. In addition, clinical metadata were not consistently available, limiting etiologic attribution of imaging patterns. Future work should incorporate more granular clinical and demographic data to enable stratified analyses by age, sex, site, and referral context, and assess AI model performance across diverse Ethiopian healthcare settings. Conclusion In the Afro-Chest X-ray multicenter Ethiopian cohort, the most frequent annotated chest X-ray abnormalities were reticulonodular or ground-glass opacities, pleural effusion, cardiomegaly, fibrotic or fibrobronchiectatic change, and consolidation or cavitation. The predominance of reticulonodular and ground-glass opacities reflect the high prevalence of infectious lung disease, particularly tuberculosis (TB) and other viral and bacterial pneumonias. These findings characterize the dominant abnormal chest X-ray patterns encountered in Ethiopian clinical practice and establish a radiologist-verified reference cohort to support regionally relevant epidemiologic description and the development and validation of AI systems. Declarations Data Availability Deidentified Afro-Chest X-ray data will be made available on reasonable request, subject to ethical approvals and institutional data governance requirements. Funding Supported by the Lacuna Fund. The funder had no role in the design or conduct of the study. Disclosures The authors declare no relevant conflicts of interest. References Wang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM. ChestX-ray8: Hospital-scale chest X-ray database and benchmarks on weakly supervised classification and localization of common thoracic diseases. Proc IEEE Conf Comput Vis Pattern Recognit . 2017:3462–3471. doi:10.1109/CVPR.2017.369 Johnson AEW, Pollard TJ, Berkowitz SJ, et al. MIMIC-chest X-ray, a de-identified publicly available database of chest radiographs with free-text reports. Sci Data . 2019;6(1):317. doi:10.1038/s41597-019-0322-0 Irvin J, Rajpurkar P, Ko M, et al. CheXpert: A large chest radiograph dataset with uncertainty labels and expert comparison. Proc AAAI Conf Artif Intell . 2019;33(1):590–597. doi:10.1609/aaai.v33i01.3301590 Nguyen HQ, Lam K, Le LT, et al. VinDr-chest X-ray: An open dataset of chest X-rays with radiologist annotations. Sci Data . 2022;9:429. doi:10.1038/s41597-022-01498-w Bustos A, Pertusa A, Salinas JM, de la Iglesia-Vayá M. PadChest: A large chest X-ray image dataset with multi-label annotated reports. Med Image Anal . 2020;66:101797. doi:10.1016/j.media.2020.101797 Mekonen KA, Mohammed SH, Kebede T, Bedane A, Buser AA. Artificial intelligence in radiology for Ethiopia. Radiology . 2024. PubMed PMID: 39735519 World Health Organization. Cardiovascular diseases (CVDs). Updated July 31, 2025. Accessed December 18, 2025. World Health Organization. The top 10 causes of death. Updated August 7, 2024. Accessed December 18, 2025. Zech JR, Badgeley MA, Liu M, Costa AB, Titano JJ, Oermann EK. Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: A cross-sectional study. PLoS Med . 2018;15(11):e1002683. doi:10.1371/journal.pmed.1002683 Mollura DJ, Culp MP, Pollack E, et al. Artificial intelligence in low- and middle-income countries: Innovating global health radiology. Radiology . 2020;297(3):513–520. doi:10.1148/radiol.2020201434 Arora A, Rajpurkar P, Hoffman J, et al. The value of standards for health datasets in artificial intelligence. NPJ Digit Med . 2023;6:210. Training and practice of radiology in Ethiopia: Challenges and prospects. Ethiop J Health Sci . 2022. PMCID: PMC9624104 Kebede T, Zewdeneh D, Atnafu A, et al. The journey of radiology in Ethiopia. Ethiop J Health Sci . 2022;32(Spec Issue 1). PubMed PMID: 36339960 World Health Organization. Patient safety: 10 facts on patient safety—medical exposure to radiation. Published August 26, 2019. Accessed December 18, 2025. Dutta A, Zisserman A. The VIA annotation software for images, audio, and video. arXiv . 2019;1904.10699 Tables Table 1. Participating institutions contributing Afro-Chest X-ray radiographs Institution City/Region Sector Notes Menelik II Comprehensive Specialized Hospital Addis Ababa Public Tertiary referral hospital Myungsung Christian Medical Center [MCM] Addis Ababa Private Teaching and referral hospital Bethzatha General Hospital Addis Ababa Private General hospital BMY Diagnostic Center Addis Ababa Private Diagnostic imaging center Wudassie Diagnostic Center Addis Ababa Private Diagnostic imaging center Viva Diagnostic Center Addis Ababa Private Diagnostic Imaging center Vision Diagnostic Center Addis Ababa Privatec Diagnostic imaging center Teklehaimanot General Hospital Addis Ababa Private General hospital Hallelujah General Hospital Addis Ababa Private General hospital Ayder Comprehensive Specialized Hospital Mekelle (Tigray) Public Teaching/referral hospital Table 2. Afro-Chest X-ray cohort assembly and analytic subsets. Dataset component Count Notes Raw DICOM radiographs retrieved 61,000 From PACS archives, before quality filtering Quality-filtered radiographs retained 55,409 Deidentified chest X-rays suitable for analysis Radiographs with paired free-text reports 31,983 Collected across institutions Matched report–image pairs 31,939 Linked to 48,962 unique patients Radiographs manually annotated 11,880 Bounding boxes for 19 findings Abnormal annotated radiographs 7,003 Contain ≥1 labeled finding Total labeled findings (instances) 22,531 Multiple findings per radiograph are possible Table 3. Afro-Chest X-ray 19-label taxonomy used for manual annotation (operational definitions summarized from the labeling guide). Label ID Finding label Operational definition 0 Reticulonodular or ground-glass opacity Diffuse or patchy interstitial or ground-glass opacity pattern; includes reticular, nodular, or mixed GGO patterns. 1 Pleural effusion Blunting of the costophrenic angle or dependent pleural fluid collection; unilateral or bilateral. 2 Cardiomegaly or chamber enlargement Enlarged cardiac silhouette or chamber enlargement on a PA or frontal radiograph. 3 Fibrosis or fibrobronchiectatic change Linear scarring, volume loss, traction bronchiectasis, or chronic fibrotic changes. 4 Consolidation or cavitary lesion Airspace consolidation and/or cavitation suspicious for necrotizing infection or tuberculosis-spectrum disease. 5 Nodule Discrete pulmonary nodule(s) or focal rounded opacity size between 7 mm – 3 cm. 6 Mediastinal or vascular abnormality Abnormal mediastinal contour, mediastinal widening or vascular enlargement. 7 Supporting device Iatrogenic device (e.g., tubes, lines, drains, pacemaker leads). 8 Pleural thickening or calcification Pleural plaque, thickening, or calcification. 9 Diaphragmatic or subdiaphragmatic abnormality Elevated hemidiaphragm, abnormal contour, or subdiaphragmatic abnormality. 10 Atelectasis Segmental or lobar volume loss with associated opacification. 11 Thoracic cage or soft tissue abnormality Rib, clavicle, or vertebral lesion or chest wall soft-tissue abnormality. 12 Bronchial or airway abnormality Central airway abnormality or airway wall thickening pattern. 13 Hilar lesion Hilar enlargement, mass, or abnormal contour. 14 Pneumothorax Pleural air with visible pleural line; includes tension physiology if evident. 15 Lung cyst Cystic lucency or bulla pattern as defined by the label guide. 16 Mass Mass-like opacity larger than 3 cm. 17 Opaque hemithorax Near-complete opacification of one hemithorax. 18 Hernia Hiatal or diaphragmatic hernia on chest X-ray. Table 4. Most frequent abnormalities in abnormal annotated radiographs (finding-instance level). Finding label Finding instances (n) % of all labeled findings (N=22,531) Reticulonodular or ground-glass opacities 6,896 30.6% Pleural effusion 2,685 11.9% Cardiomegaly or chamber enlargement 2,161 9.6% Fibrosis or fibrobronchiectatic change 1,869 8.3% Consolidation or cavitary lesion 1,835 8.1% Table 5. Afro Chest X-ray template ID Section Normal Finding Abnormal Options Mapped Annotations (ID) 1 Trachea Midline and patent, normal airways Deviation, narrowing, mass effect Bronchial/Airway abnormalities (ID 15) 2 Hilar Structures Normal size and contour Enlargement, mass, abnormal contour Hilar lesions (ID 6) 3 Lung Fields Right Lung: Clear; Left Lung: Clear Opacity, volume loss, cystic changes Atelectasis (4), Pleural thickening (5), Fibrosis (7), Opaque hemithorax (14), Lung cyst (17) 5 Parenchymal Consolidation Not present Present — location specified Consolidation/Cavitary lesion (ID 0) 6 Reticulonodular/Interstitial Not present Present — pattern specified Reticulonodular/GGO (ID 1) 7 Nodules or Masses Not present Present — size, location noted Nodule (10), Mass (12) 4 Heart & Mediastinum Normal size and contour Cardiomegaly, mediastinal widening Cardiomegaly (2), Mediastinal/Vascular (9) 8 Pleural Effusion Not present Present — laterality, size noted Pleural effusion (ID 3) 9 Pneumothorax Not present Present — size, tension noted Pneumothorax (ID 8) 10 Diaphragm Normal position and contour bilaterally; Sharp costophrenic angles Elevated, flattened, hernia, subdiaphragmatic air Diaphragmatic abnormality (11), Opaque hemithorax (14), Hernia (16) 11 Bones & Soft Tissues Ribs: No fracture; Spine: Normal; Clavicles: Intact; Soft tissues: Unremarkable Fracture, lesion, soft tissue mass Thoracic cage/Soft tissue (ID 13) 12 Supporting Devices No lines, tubes, or devices present Central line, ET tube, NG tube, Pacemaker, Chest tube — position noted Supporting devices (ID 18) T/emplate_02 1: 'Trachea: Midline and patent, normal airways', 2: 'Hilar Structures: Normal size and contour', 3: 'Lung Fields: \ n \ Right Lung: Clear\n\ Left Lung: Clear', 4: 'Heart size and mediastinal outline: Normal', 5: 'Parenchymal consolidation: yes/no, 6: 'Reticulonodular opacity/ Interstitial Markings: yes/no 7: 'Nodules or Masses: yes/no, 8: 'Pleural Effusion: yes/no, 9: 'Pneumothorax: yes/no, 10: 'Diaphragm\n\ Right Hemidiaphragm: Normal position and contour\n\ Left Hemidiaphragm: Normal position and contour\n\ Costophrenic Angles: Sharp and clear\n\ Subdiaphragmatic Air/Abnormality: yes/no, 11: 'Bones and Soft Tissues:\n\ Ribs: No fracture or lesion\n\ Spine: No abnormal curvature or compression\n\ Clavicles: Intact\n\ Shoulder Girdle: Unremarkable\n\ Soft Tissues: No swelling or subcutaneous air 12: 'Supporting Lines/Tubes/Devices:\n\ Central Line: Not present\n\ ET Tube: Not present\n\ NG Tube: Not present\n\ Pacemaker or Leads: Not present\n\ Chest Tube: Not present annotation_id2label = { 0: "Consolidation or Cavitory lesion", 1: "Reticulonodular and ground glass opacity", 2: "Cardiomegaly or chamber enlargement", 3: "Pleural effusion", 4: "Atelectasis", 5: "Pleural thickening or Calcification", 6: "Hilar lesions", 7: "Fibrosis and fibrobronchiectatic changes", 8: "Pneumothorax", 9: "Mediastinal or Vascular and Aortic lesions", 10: "Nodule", 11: "Diaphragmatic or subdiaphragmatic abnormalities", 12: "Mass", 13: "Thoracic- cage or Soft tissue or Spinal abnormalities", 14: "Opaque hemithorax", 15: "Bronchial or Airway abnormalities", 16: "Hernia", 17: "Lung cyst or cystic lesions", 18: "Supporting devices"} How we matched between the annotation ID with the report template template_02_id2annotation_id = { 1 : [15], 2 : [6], 3 : [ 4,5, 7, 14, 17], 4 : [2, 9], 5 : [0], 6 : [1], 7 : [10, 12], 8 : [3], 9 : [8], 10: [ 11, 14, 16], 11: [13], 12: [18]} annotation_id2template_02_id = { 0 : [5], 1 : [6], 2 : [4], 3 : [8], 4 : [3], 5 : [3], 6 : [2], 7 : [3], 8 : [9], 9 : [4], 10: [7], 11: [10], 12: [7], 13: [11], 14: [3,10], 15: [1], 16: [10], 17: [3], 18: [12]} Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9417593","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":623108856,"identity":"6f1e56af-146f-4d4b-b42a-78db6a5c2c79","order_by":0,"name":"Messay Gebrekidan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABAElEQVRIiWNgGAWjYBCDBAYG5gMHEiqATGbmBmK1sCUeeHAGpIWRaC08xgcftoHYBLTotrc//MzzyybPvL3H4EDivNpo/naglh8V23BqMTtzxliaty+tWObMsYIDiduO5844zNjA2HPmNm4tN3IYpHl7DifOkEjeANRyLLcBqIWZsQ2PlvvPH//m7fmfOEP+AdBhc47lzieo5QaDmTTPjwNAW1iAWhpqcjcQ1HImx8xybkNysQRPWsKBhGMHcjcCtRzE65fjxx/fePPHLk+C/fDhjz9q6nLnnT988MGPCtxaQICJtw3OPgwmD+BVDwSMP/7A2XWEFI+CUTAKRsEIBAAQumjhdsAAzgAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-6698-1321","institution":"TiruRad diagnostics and research center","correspondingAuthor":true,"prefix":"","firstName":"Messay","middleName":"","lastName":"Gebrekidan","suffix":""},{"id":623108857,"identity":"c3b2a5e7-14a7-4996-964b-401b7ee97168","order_by":1,"name":"Samuel S. Hailu","email":"","orcid":"https://orcid.org/0000-0001-5543-3215","institution":"Department of Radiology, Children's Hospital of Philadelphia, Philadelphia, PA 19104, USA","correspondingAuthor":false,"prefix":"","firstName":"Samuel","middleName":"S.","lastName":"Hailu","suffix":""},{"id":623108858,"identity":"d5d8d20c-0d06-492b-9f67-acc4ec7b7ee1","order_by":2,"name":"Michael Teklehaimanot Abera","email":"","orcid":"https://orcid.org/0009-0003-0241-2965","institution":"Addis Ababa University College of health sciences, Addis Ababa,Ethiopia","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"Teklehaimanot","lastName":"Abera","suffix":""},{"id":623108859,"identity":"0914e5cb-7917-4a77-b64c-2d6478f8eade","order_by":3,"name":"Melkamu Hunegnaw Asmare","email":"","orcid":"https://orcid.org/0000-0003-0990-3348","institution":"Leuven Center for Affordable Health Technology, KU Leuven, Andreas Vesaliusstraat 13, Leuven, 3000, Flanders, Belgium","correspondingAuthor":false,"prefix":"","firstName":"Melkamu","middleName":"Hunegnaw","lastName":"Asmare","suffix":""},{"id":623108860,"identity":"aa924b6e-03f6-4f42-91e2-5546028b9d83","order_by":4,"name":"Michael A. Negussie","email":"","orcid":"https://orcid.org/0000-0001-9095-5668","institution":"School of Medicine, College of Health Sciences, Addis Ababa University, Addis Ababa, Ethiopia","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"A.","lastName":"Negussie","suffix":""}],"badges":[],"createdAt":"2026-04-14 15:48:51","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":true,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9417593/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9417593/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107258763,"identity":"b528e27b-bbe0-4daa-9028-9cd8d2c02ed8","added_by":"auto","created_at":"2026-04-19 12:40:18","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":169101,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eThe flow diagram shows cohort assembly and the analytic subset used to summarize radiographic finding patterns.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/2bb70ef8efb530317ebff976.png"},{"id":107258768,"identity":"4e8ce7d3-6203-4821-8297-6790dda750a2","added_by":"auto","created_at":"2026-04-19 12:40:19","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":54281,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eGeographic distribution of participating Afro-Chest X-ray sites in Ethiopia (schematic map).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/3e6e69da22d2576ee0945a62.png"},{"id":107258753,"identity":"9dbfae78-bf01-4ec4-b277-61afd102541f","added_by":"auto","created_at":"2026-04-19 12:40:17","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":96205,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDistribution of the most frequent annotated abnormalities (counts and percentage of all labeled findings).\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/91dc844f54c0c6dbac0bb91b.png"},{"id":107258767,"identity":"332f1a7a-8b67-4461-9321-6655ca642862","added_by":"auto","created_at":"2026-04-19 12:40:19","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":163833,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eAfro-Chest X-ray clinical workflow from multicenter acquisition to expert labels.\u003c/strong\u003e\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/07125f54bb749ec856842446.png"},{"id":107258765,"identity":"168b8e2a-d7f9-4b79-8608-8cb4b6d6f689","added_by":"auto","created_at":"2026-04-19 12:40:18","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":581267,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eExamples of Afro-Chest X-ray manual labeling.\u003c/strong\u003e\u003cbr\u003e\n(A) \u003cstrong\u003eNo findings\u003c/strong\u003e example labeled as normal without bounding boxes. (B) Example abnormal radiograph demonstrating a \u003cstrong\u003erectangular bounding box\u003c/strong\u003eindicating the finding location and the \u003cstrong\u003elabeler's confidence\u003c/strong\u003e recorded at annotation time. Confidence values reflect the radiologist’s self-rated certainty using the Afro-Chest X-ray labeling rubric.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/a915f44a279cc55e222b0f97.png"},{"id":107258787,"identity":"2f5ccef4-48d0-4c27-882b-f7c9d0efeac2","added_by":"auto","created_at":"2026-04-19 12:40:25","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":1217589,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRepresentative uncommon thoracic entities in the Afro-Chest X-ray multicenter cohort (Ethiopia).\u003c/strong\u003e\u003cbr\u003e\nPanels show examples of selected \u003cstrong\u003eless frequent entities\u003c/strong\u003e with Afro-Chest X-ray rectangular bounding boxes indicating lesion localization and the recorded labeler confidence. (A) Diffuse bilateral lung nodules secondary to metastasis. (B) Left lower lung zone water lily sign, indicating a contained rupture of a left lower lung hydatid cyst. (C) Right lower lung radio-opaque lung mass. (D) A large right lower lung-centered radio-opaque lesion with sharp borders representing an uncomplicated lung hydatid cyst. (E) Proximal esophageal dilatation with an air-fluid level, a case of achalasia. (F) A case of congenital lobar hyperinflation with associated right lung atelectasis in a neonate.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/f0dffda177f0e68c735c37f1.png"},{"id":107258756,"identity":"2ac41146-0454-4232-8230-a98a4220991e","added_by":"auto","created_at":"2026-04-19 12:40:17","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1064113,"visible":true,"origin":"","legend":"\u003cp\u003eSample raw chest X-ray images\u003c/p\u003e","description":"","filename":"image13.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/e13b38ac041428ab573cf970.png"},{"id":107258775,"identity":"8d192861-0b9e-48bf-8e60-dbff1ce76081","added_by":"auto","created_at":"2026-04-19 12:40:22","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":236711,"visible":true,"origin":"","legend":"\u003cp\u003eHow a normal chest X-ray is annotated as no findings.\u003c/p\u003e","description":"","filename":"image14.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/5eb7dce159b2343ffe37fd65.png"},{"id":107258789,"identity":"83ba24c5-913a-4363-b48c-a356447b71eb","added_by":"auto","created_at":"2026-04-19 12:40:26","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":232683,"visible":true,"origin":"","legend":"\u003cp\u003eHow the right upper lobe consolidation is labeled with a high confidence interval and marked as pneumonia\u003c/p\u003e","description":"","filename":"image15.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/89bfe8986c98860e68c9495f.png"},{"id":107484812,"identity":"3745e7b4-5fac-47c6-a57f-60f78d8e52f2","added_by":"auto","created_at":"2026-04-22 02:33:03","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":260255,"visible":true,"origin":"","legend":"\u003cp\u003eCardiomegaly with the bounding box.\u003c/p\u003e","description":"","filename":"image16.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/81934c78e9b5b4a48f69f7f6.png"},{"id":107258774,"identity":"a3714ec9-b375-4b6f-94bd-afd503dce542","added_by":"auto","created_at":"2026-04-19 12:40:22","extension":"png","order_by":11,"title":"Figure 11","display":"","copyAsset":false,"role":"figure","size":281187,"visible":true,"origin":"","legend":"\u003cp\u003eBilateral mid and lower central lung zones post-infectious bronchiectatic changes\u003c/p\u003e","description":"","filename":"image17.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/da6ab7ea541dd8872ef5c2da.png"},{"id":107258785,"identity":"9c18bd7f-006f-4510-a69d-e414e86d098d","added_by":"auto","created_at":"2026-04-19 12:40:24","extension":"png","order_by":12,"title":"Figure 12","display":"","copyAsset":false,"role":"figure","size":393901,"visible":true,"origin":"","legend":"\u003cp\u003eRight moderate pleural effusion labelled\u003c/p\u003e","description":"","filename":"image18.png","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/2b04b50ad5ef33afd4b865b4.png"},{"id":107705351,"identity":"0c0c24c6-5d9a-4ec0-b75b-c0aea99dc456","added_by":"auto","created_at":"2026-04-24 09:11:44","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":5081395,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9417593/v1/73c42f20-9e0c-4091-93ac-5aec91bcd58f.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eMulticenter Chest Radiographic Findings from Ethiopia with Radiologist-Generated Annotations\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Advances in Knowledge","content":"\u003cul\u003e\n \u003cli\u003eThis multicenter Ethiopian cohort demonstrates that reticulonodular or ground-glass opacities, pleural effusion, cardiomegaly, and fibrotic changes account for most chest X-ray abnormalities in routine clinical practice.\u003c/li\u003e\n \u003cli\u003eThe Afro-Chest X-ray provides a large, radiologist-annotated reference dataset from an under-represented East African population, enabling characterization of regional chest radiographic patterns and supporting improved generalizability of imaging research and AI systems beyond high-income settings.\u003c/li\u003e\n\u003c/ul\u003e\n\u003ch2\u003eImplications for Patient Care\u003c/h2\u003e\n\u003cul\u003e\n \u003cli\u003eRegion-specific chest X-ray patterns underscore the need for local validation before deployment of AI interpretation tools to ensure accurate and safe use in patient care.\u003c/li\u003e\n \u003cli\u003eA curated, radiologist-verified Ethiopian chest X-ray cohort provides a reference standard that can support local performance assessment, quality improvement, and adaptation of AI tools for high-burden cardiopulmonary conditions in East African healthcare settings.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"Introduction","content":"\u003cp\u003eChest radiography is the most accessible and widely used imaging modality in Ethiopia and is central to the evaluation of respiratory and cardiac disease. Globally, an estimated 3.6\u0026nbsp;billion diagnostic radiology examinations are performed annually, underscoring the central role of X-ray imaging in healthcare systems [14]. In parallel, AI methods for chest X-ray interpretation have advanced rapidly, largely driven by the availability of large, publicly accessible datasets primarily derived from high-income settings [1\u0026ndash;5]. The under-representation of African clinical imaging data, characterized by distinct image acquisitions and region-specific disease patterns, may reduce the generalizability of AI models [9, 10].\u003c/p\u003e \u003cp\u003eIn Ethiopia, AI-enabled radiology is emerging; however, publicly described, multicenter, chest X-ray cohorts curated by radiologists remain scarce [6]. The Afro-Chest X-ray initiative was developed to address this gap through the creation of a multicenter Ethiopian cohort with linked radiology reports and radiologist-generated localization labels for 19 findings (reference to table).\u003c/p\u003e \u003cp\u003eThe objectives of this study were to (a) describe national-level patterns of annotated chest X-ray findings in Ethiopia using the Afro-Chest X-ray cohort and (b) document the clinical workflow used to generate standardized, high-quality labels suitable for epidemiologic description and supervised AI development. This study focuses on imaging patterns and annotation methodology rather than disease prevalence or AI performance evaluation.\u003c/p\u003e \u003cp\u003eEthiopia, the second most populous nation in Africa, has approximately 600 radiology practitioners, including radiologists and radiology trainees, corresponding to roughly 0.92 radiologists per 200,000 population [13]. Similar workforce constraints are present across many African countries and contribute to high clinical workload and limited access to subspecialty expertise, with potential downstream effects on diagnostic quality, service, and radiologist burnout [10, 12]. These realities underscore the need for scalable, context-aware supportive technologies in diagnostic imaging. High-quality reference datasets that reflect local practice and disease patterns are essential to the development of such technology.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003ch2\u003eStudy Design and Ethics\u003c/h2\u003e\n\u003cp\u003eThis retrospective multicenter cohort study included deidentified chest X-rays obtained during routine clinical care at 10 Ethiopian institutions. National and institutional approvals were obtained through national and local ethics oversight pathways. Because the dataset was deidentified and collected retrospectively, informed consent was waived where applicable.\u003c/p\u003e\n\u003ch2\u003eParticipating Institutions\u003c/h2\u003e\n\u003cp\u003eParticipating institutions included a mix of public referral hospitals, teaching hospitals, and private diagnostic centers, primarily located in Addis Ababa, with additional data from a tertiary referral hospital in Mekelle (Tigray Region). This institutional composition was intended to reflect routine chest radiography practice across major Ethiopian imaging providers, while acknowledging that some geographic regions remain under-represented.\u003c/p\u003e\n\u003ch2\u003eData Acquisition and Preprocessing\u003c/h2\u003e\n\u003cp\u003eChest radiographs were exported from institution-specific PACS archives as DICOM files and underwent deidentification and quality filtering to remove non-chest images and studies unsuitable for analysis\u003ca id=\"_anchor_2\" onmouseover=\"msoCommentShow('_anchor_2','_com_2')\" onmouseout=\"msoCommentHide('_com_2')\" href=\"#_msocom_2\" language=\"JavaScript\" name=\"_msoanchor_2\"\u003e\u003c/a\u003e . Approximately 61,000 DICOM radiographs were retrieved, of which 55,409 chest X-rays were retained following preprocessing.\u003c/p\u003e\n\u003ch2\u003eRadiology Report Harmonization\u003c/h2\u003e\n\u003cp\u003eReporting practices varied across participating institutions. To enable consistent label mapping and cross-site comparison, free-text reports were normalized to the Afro-Chest X-ray reporting template with standardized sections (Technique, Findings, Impression). Report terminology was mapped to a unified 19-label taxonomy through a predefined dictionary. In internal consistency checks, concordance between standardized report-derived labels and corresponding expert annotations exceeded 90% (90.4%).\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003eExpert Annotation Protocol\u003c/h2\u003e\n\u003cp\u003eA stratified subset of radiographs was selected for expert manual annotation to enrich the representation of abnormal findings. Eleven radiologists completed structured training and calibration on the VIA annotation tool [15] prior to labeling. Radiologists annotated 19 thoracic finding categories using bounding boxes and recorded confidence scores for each finding instance. Examinations were categorized as diagnostic or nondiagnostic to reflect image interpretability. The Radiologists were given the patient\u0026apos;s gender and age for annotation, but no clinical history.\u003c/p\u003e\n\u003cp\u003eTo illustrate the structure of the manual labels, representative examples are provided of (a) a normal radiograph labeled \u0026ldquo;no findings\u0026rdquo; without bounding boxes and (b) an abnormal radiograph with bounding-box localization and an associated labeler confidence score (Figure 5). Representative uncommon entities in the cohort are shown in Figure 6, with panel-level diagnoses and key imaging findings summarized in Table S1.\u003c/p\u003e\n\u003ch2\u003eQuality Assurance\u003c/h2\u003e\n\u003cp\u003eA validation panel reviewed label definitions, adjudicated ambiguous cases, and enforced consistency of label usage across annotators. Discrepant examinations were resolved by consensus review.\u003c/p\u003e\n\u003ch2\u003eStatistical Analysis\u003c/h2\u003e\n\u003cp\u003eAnalyses were descriptive. Categorical variables are reported as counts and percentages. In the annotated subset, finding frequencies are reported at the finding-instance level, allowing multiple findings per radiograph.\u003c/p\u003e"},{"header":"Results","content":"\u003ch2\u003eCohort Characteristics\u003c/h2\u003e\n\u003cp\u003eAfter preprocessing, the Afro-Chest X-ray cohort included 55,409 chest X-rays. Of these, 31,939 radiographs were linked to radiology reports from 48,971 unique patients (male, 18,324 [37.4%]; female, 30,387 [62.1%]; unspecified sex, 260 [0.5%]). Some patients have multiple examinations, reflecting follow-up imaging to see the response to treatments performed as part of routine clinical care.\u003c/p\u003e\n\u003ch2\u003eAnnotated Subset and Abnormality Burden\u003c/h2\u003e\n\u003cp\u003eA total of 11,880 chest radiographs underwent radiologist annotation. Of these, 7,003 (58.9%) were classified as abnormal and contained 22,531 labeled finding instances (mean, 3.22 instances per abnormal radiograph). The remaining 4,877 examinations (41%) were classified as normal diagnostic examinations, and 180 radiographs were deemed non-diagnostic.\u003c/p\u003e\n\u003ch2\u003eNational Patterns of Common Findings\u003c/h2\u003e\n\u003cp\u003eThe five most frequent abnormality categories (Table 4) accounted for 15,446 of 22,531 labeled finding instances (68.6%). Reticulonodular or ground-glass opacity was the most frequently annotated finding. Pleural effusion and cardiomegaly were also common, and chronic fibrotic change and consolidation or cavitation comprised substantial portions of abnormal labels. Less frequently annotated findings included hernia, opaque hemithorax, mass, lung cyst, pneumothorax, and diaphragmatic or subdiaphragmatic abnormalities.\u003c/p\u003e\n\u003cp\u003eBeyond common abnormalities, the Afro-Chest X-ray cohort contains uncommon and regionally important entities that are often underrepresented in widely used public chest X-ray datasets. Figure 6 illustrates representative examples, including pulmonary hydatid disease patterns (including a \u0026ldquo;water lily\u0026rdquo; configuration), diffuse metastatic nodules, a lung mass, esophageal dilation compatible with achalasia, and congenital lobar hyperinflation. These examples complement the national-level prevalence findings by highlighting the cohort\u0026rsquo;s phenotypic breadth and clinical diversity.\u003c/p\u003e\u003cp\u003eFigure 1 summarizes cohort assembly.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFigure 2 shows site distribution.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFigure 3 shows the most frequent annotated findings.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFigure 4 illustrates the Afro-Chest X-ray workflow.\u003c/p\u003e\n\u003cp\u003eFigure 5Examples of Afro-Chest X-ray manual labeling.\u003c/p\u003e\n\u003cp\u003eFigure 6 Unique and uncommon entities (multi-panel)\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this multicenter Ethiopian cohort, annotated chest X-ray abnormalities were dominated by parenchymal opacities, pleural effusion, and cardiomegaly. This distribution is consistent with routine clinical imaging indications in Ethiopia, where infectious and inflammatory lung disease, pleural pathology, and cardiac conditions account for a substantial proportion of chest X-ray utilization [7, 8]. The prominence of fibrotic and fibrobronchiectatic changes further suggests that chronic or post-infectious lung disease is frequently encountered on chest X-rays interpreted across Ethiopian institutions.\u003c/p\u003e \u003cp\u003eThese radiographic findings have practical implications for both clinical training and AI development. First, they provide a data-driven characterization of the most common abnormality patterns encountered in participating facilities. Second, they highlight class-imbalance considerations relevant to AI development: while the most prevalent imaging patterns support robust model training, less common but clinically important findings may require targeted sampling strategies or collaborative data sharing to ensure adequate representation [11]. The Afro-Chest X-ray dataset was designed to support both classification and localization tasks by using radiologist-generated bounding box confidence scoring and a consensus-based quality assurance process [15]. To enable scalable annotation across institutions with heterogeneous reporting practices, free-text radiology reports were standardized to a shared reporting template. The labeling process was iterative and designed to support comprehensive abnormality detection. Annotations distinguished critical findings, such as active consolidation, from subcritical or chronic abnormalities. Label normalization incorporated finding-level prioritization, emphasizing clinically significant selected findings (e.g., pneumothorax, atelectasis, and pulmonary nodule).\u003c/p\u003e \u003cp\u003eLimitations\u003c/p\u003e \u003cp\u003eThis study has limitations. Most chest radiographs were collected from institutions in Addis Ababa, with additional data from Mekelle, resulting in limited direct representation of other regions in Ethiopia. However, Addis Ababa serves as the country's largest national referral hub, receiving patients from across Ethiopia and neighboring East African countries (including Sudan, Eritrea, and Djibouti). Consequently, despite the geographic concentration of participating sites, the cohort likely captures a broad spectrum of national and regional disease patterns relevant to East African clinical practice.\u003c/p\u003e \u003cp\u003eIn addition, clinical metadata were not consistently available, limiting etiologic attribution of imaging patterns. Future work should incorporate more granular clinical and demographic data to enable stratified analyses by age, sex, site, and referral context, and assess AI model performance across diverse Ethiopian healthcare settings.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn the Afro-Chest X-ray multicenter Ethiopian cohort, the most frequent annotated chest X-ray abnormalities were reticulonodular or ground-glass opacities, pleural effusion, cardiomegaly, fibrotic or fibrobronchiectatic change, and consolidation or cavitation. The predominance of reticulonodular and ground-glass opacities reflect the high prevalence of infectious lung disease, particularly tuberculosis (TB) and other viral and bacterial pneumonias. These findings characterize the dominant abnormal chest X-ray patterns encountered in Ethiopian clinical practice and establish a radiologist-verified reference cohort to support regionally relevant epidemiologic description and the development and validation of AI systems.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eData Availability\u003c/h2\u003e\n\u003cp\u003eDeidentified Afro-Chest X-ray data will be made available on reasonable request, subject to ethical approvals and institutional data governance requirements.\u003c/p\u003e\n\u003ch2\u003eFunding\u003c/h2\u003e\n\u003cp\u003eSupported by the Lacuna Fund. The funder had no role in the design or conduct of the study.\u003c/p\u003e\n\u003ch2\u003eDisclosures\u003c/h2\u003e\n\u003cp\u003eThe authors declare no relevant conflicts of interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eWang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM. ChestX-ray8: Hospital-scale chest X-ray database and benchmarks on weakly supervised classification and localization of common thoracic diseases. \u003cem\u003eProc IEEE Conf Comput Vis Pattern Recognit\u003c/em\u003e. 2017:3462\u0026ndash;3471. doi:10.1109/CVPR.2017.369\u003c/li\u003e\n\n\u003cli\u003eJohnson AEW, Pollard TJ, Berkowitz SJ, et al. MIMIC-chest X-ray, a de-identified publicly available database of chest radiographs with free-text reports. \u003cem\u003eSci Data\u003c/em\u003e. 2019;6(1):317. doi:10.1038/s41597-019-0322-0\u003c/li\u003e\n\n\u003cli\u003eIrvin J, Rajpurkar P, Ko M, et al. CheXpert: A large chest radiograph dataset with uncertainty labels and expert comparison. \u003cem\u003eProc AAAI Conf Artif Intell\u003c/em\u003e. 2019;33(1):590\u0026ndash;597. doi:10.1609/aaai.v33i01.3301590\u003c/li\u003e\n\n\u003cli\u003eNguyen HQ, Lam K, Le LT, et al. VinDr-chest X-ray: An open dataset of chest X-rays with radiologist annotations. \u003cem\u003eSci Data\u003c/em\u003e. 2022;9:429. doi:10.1038/s41597-022-01498-w\u003c/li\u003e\n\n\u003cli\u003eBustos A, Pertusa A, Salinas JM, de la Iglesia-Vay\u0026aacute; M. PadChest: A large chest X-ray image dataset with multi-label annotated reports. \u003cem\u003eMed Image Anal\u003c/em\u003e. 2020;66:101797. doi:10.1016/j.media.2020.101797\u003c/li\u003e\n\n\u003cli\u003eMekonen KA, Mohammed SH, Kebede T, Bedane A, Buser AA. Artificial intelligence in radiology for Ethiopia. \u003cem\u003eRadiology\u003c/em\u003e. 2024. PubMed PMID: 39735519\u003c/li\u003e\n\n\u003cli\u003eWorld Health Organization. Cardiovascular diseases (CVDs). Updated July 31, 2025. Accessed December 18, 2025.\u003c/li\u003e\n\n\u003cli\u003eWorld Health Organization. The top 10 causes of death. Updated August 7, 2024. Accessed December 18, 2025.\u003c/li\u003e\n\n\u003cli\u003eZech JR, Badgeley MA, Liu M, Costa AB, Titano JJ, Oermann EK. Variable generalization performance of a deep learning model to detect pneumonia in chest radiographs: A cross-sectional study. \u003cem\u003ePLoS Med\u003c/em\u003e. 2018;15(11):e1002683. doi:10.1371/journal.pmed.1002683\u003c/li\u003e\n\n\u003cli\u003eMollura DJ, Culp MP, Pollack E, et al. Artificial intelligence in low- and middle-income countries: Innovating global health radiology. \u003cem\u003eRadiology\u003c/em\u003e. 2020;297(3):513\u0026ndash;520. doi:10.1148/radiol.2020201434\u003c/li\u003e\n\n\u003cli\u003eArora A, Rajpurkar P, Hoffman J, et al. The value of standards for health datasets in artificial intelligence. \u003cem\u003eNPJ Digit Med\u003c/em\u003e. 2023;6:210.\u003c/li\u003e\n\n\u003cli\u003eTraining and practice of radiology in Ethiopia: Challenges and prospects. \u003cem\u003eEthiop J Health Sci\u003c/em\u003e. 2022. PMCID: PMC9624104 \u003c/li\u003e\n\n\u003cli\u003eKebede T, Zewdeneh D, Atnafu A, et al. The journey of radiology in Ethiopia. \u003cem\u003eEthiop J Health Sci\u003c/em\u003e. 2022;32(Spec Issue 1). PubMed PMID: 36339960\u003c/li\u003e\n\n\u003cli\u003eWorld Health Organization. Patient safety: 10 facts on patient safety\u0026mdash;medical exposure to radiation. Published August 26, 2019. Accessed December 18, 2025.\u003c/li\u003e\n\n\u003cli\u003eDutta A, Zisserman A. The VIA annotation software for images, audio, and video. \u003cem\u003earXiv\u003c/em\u003e. 2019;1904.10699\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003e\u003cstrong\u003eTable 1. Participating institutions contributing Afro-Chest X-ray radiographs\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"575\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInstitution\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCity/Region\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSector\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNotes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eMenelik II Comprehensive Specialized Hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePublic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eTertiary referral hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eMyungsung Christian Medical Center [MCM]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eTeaching and referral hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eBethzatha General Hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eGeneral hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eBMY Diagnostic Center\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eDiagnostic imaging center\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eWudassie Diagnostic Center\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eDiagnostic imaging center\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eViva Diagnostic Center\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eDiagnostic Imaging center\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eVision Diagnostic Center\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivatec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eDiagnostic imaging center\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eTeklehaimanot General Hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eGeneral hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eHallelujah General Hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAddis Ababa\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePrivate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eGeneral hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eAyder Comprehensive Specialized Hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 144px;\"\u003e\n \u003cp\u003eMekelle (Tigray)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePublic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003eTeaching/referral hospital\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2. Afro-Chest X-ray cohort assembly and analytic subsets.\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"575\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDataset component\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCount\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNotes\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eRaw DICOM radiographs retrieved\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e61,000\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eFrom PACS archives, before quality filtering\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eQuality-filtered radiographs retained\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e55,409\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eDeidentified chest X-rays suitable for analysis\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eRadiographs with paired free-text reports\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e31,983\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eCollected across institutions\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eMatched report\u0026ndash;image pairs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e31,939\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eLinked to 48,962 unique patients\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eRadiographs manually annotated\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e11,880\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eBounding boxes for 19 findings\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eAbnormal annotated radiographs\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e7,003\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eContain \u0026ge;1 labeled finding\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eTotal labeled findings (instances)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 108px;\"\u003e\n \u003cp\u003e22,531\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 258px;\"\u003e\n \u003cp\u003eMultiple findings per radiograph are possible\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3. Afro-Chest X-ray 19-label taxonomy used for manual annotation (operational definitions summarized from the labeling guide).\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"575\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLabel ID\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFinding label\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eOperational definition\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eReticulonodular or ground-glass opacity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eDiffuse or patchy interstitial or ground-glass opacity pattern; includes reticular, nodular, or mixed GGO patterns.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003ePleural effusion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eBlunting of the costophrenic angle or dependent pleural fluid collection; unilateral or bilateral.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eCardiomegaly or chamber enlargement\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eEnlarged cardiac silhouette or chamber enlargement on a PA or frontal radiograph.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eFibrosis or fibrobronchiectatic change\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eLinear scarring, volume loss, traction bronchiectasis, or chronic fibrotic changes.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eConsolidation or cavitary lesion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eAirspace consolidation and/or cavitation suspicious for necrotizing infection or tuberculosis-spectrum disease.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eNodule\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eDiscrete pulmonary nodule(s) or focal rounded opacity size between 7 mm \u0026ndash; 3 cm.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eMediastinal or vascular abnormality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eAbnormal mediastinal contour, mediastinal widening or vascular enlargement.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eSupporting device\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eIatrogenic device (e.g., tubes, lines, drains, pacemaker leads).\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003ePleural thickening or calcification\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003ePleural plaque, thickening, or calcification.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eDiaphragmatic or subdiaphragmatic abnormality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eElevated hemidiaphragm, abnormal contour, or subdiaphragmatic abnormality.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eAtelectasis\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eSegmental or lobar volume loss with associated opacification.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eThoracic cage or soft tissue abnormality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eRib, clavicle, or vertebral lesion or chest wall soft-tissue abnormality.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eBronchial or airway abnormality\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eCentral airway abnormality or airway wall thickening pattern.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eHilar lesion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eHilar enlargement, mass, or abnormal contour.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003ePneumothorax\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003ePleural air with visible pleural line; includes tension physiology if evident.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eLung cyst\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eCystic lucency or bulla pattern as defined by the label guide.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eMass\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eMass-like opacity larger than 3 cm.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eOpaque hemithorax\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eNear-complete opacification of one hemithorax.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 47px;\"\u003e\n \u003cp\u003e18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 207px;\"\u003e\n \u003cp\u003eHernia\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 321px;\"\u003e\n \u003cp\u003eHiatal or diaphragmatic hernia on chest X-ray.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4. Most frequent abnormalities in abnormal annotated radiographs (finding-instance level).\u003c/strong\u003e\u003c/p\u003e\n\u003cdiv align=\"\"\u003e\n \u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"575\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFinding label\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eFinding instances (n)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e% of all labeled findings (N=22,531)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eReticulonodular or ground-glass opacities\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003e6,896\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e30.6%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003ePleural effusion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003e2,685\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e11.9%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eCardiomegaly or chamber enlargement\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003e2,161\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e9.6%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eFibrosis or fibrobronchiectatic change\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003e1,869\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e8.3%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 210px;\"\u003e\n \u003cp\u003eConsolidation or cavitary lesion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 174px;\"\u003e\n \u003cp\u003e1,835\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 192px;\"\u003e\n \u003cp\u003e8.1%\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTable 5. Afro Chest X-ray template\u0026nbsp;\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"526\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eID\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSection\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNormal Finding\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAbnormal Options\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMapped Annotations (ID)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTrachea\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eMidline and patent, normal airways\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eDeviation, narrowing, mass effect\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eBronchial/Airway abnormalities (ID 15)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHilar Structures\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNormal size and contour\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eEnlargement, mass, abnormal contour\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eHilar lesions (ID 6)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLung Fields\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eRight Lung: Clear; Left Lung: Clear\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eOpacity, volume loss, cystic changes\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eAtelectasis (4), Pleural thickening (5), Fibrosis (7), Opaque hemithorax (14), Lung cyst (17)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e5\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eParenchymal Consolidation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNot present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003ePresent \u0026mdash; location specified\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eConsolidation/Cavitary lesion (ID 0)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e6\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eReticulonodular/Interstitial\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNot present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003ePresent \u0026mdash; pattern specified\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eReticulonodular/GGO (ID 1)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e7\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNodules or Masses\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNot present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003ePresent \u0026mdash; size, location noted\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eNodule (10), Mass (12)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e4\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHeart \u0026amp; Mediastinum\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNormal size and contour\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eCardiomegaly, mediastinal widening\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eCardiomegaly (2), Mediastinal/Vascular (9)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e8\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePleural Effusion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNot present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003ePresent \u0026mdash; laterality, size noted\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003ePleural effusion (ID 3)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e9\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003ePneumothorax\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNot present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003ePresent \u0026mdash; size, tension noted\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003ePneumothorax (ID 8)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e10\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDiaphragm\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNormal position and contour bilaterally; Sharp costophrenic angles\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eElevated, flattened, hernia, subdiaphragmatic air\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eDiaphragmatic abnormality (11), Opaque hemithorax (14), Hernia (16)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e11\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eBones \u0026amp; Soft Tissues\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eRibs: No fracture; Spine: Normal; Clavicles: Intact; Soft tissues: Unremarkable\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eFracture, lesion, soft tissue mass\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eThoracic cage/Soft tissue (ID 13)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 32px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e12\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 158px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSupporting Devices\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 97px;\"\u003e\n \u003cp\u003eNo lines, tubes, or devices present\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 106px;\"\u003e\n \u003cp\u003eCentral line, ET tube, NG tube, Pacemaker, Chest tube \u0026mdash; position noted\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eSupporting devices (ID 18)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eT/emplate_02\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 1: \u0026apos;Trachea: Midline and patent, normal airways\u0026apos;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 2: \u0026apos;Hilar Structures: Normal size and contour\u0026apos;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 3: \u0026apos;Lung Fields: \\ n \\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Right Lung: Clear\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Left Lung: Clear\u0026apos;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 4: \u0026apos;Heart size and mediastinal outline: Normal\u0026apos;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 5: \u0026apos;Parenchymal consolidation: yes/no,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 6: \u0026apos;Reticulonodular \u0026nbsp;opacity/ Interstitial Markings: yes/no\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 7: \u0026apos;Nodules or Masses: yes/no,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 8: \u0026apos;Pleural Effusion: yes/no,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 9: \u0026apos;Pneumothorax: yes/no,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 10: \u0026apos;Diaphragm\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Right Hemidiaphragm: Normal position and contour\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Left Hemidiaphragm: Normal position and contour\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Costophrenic Angles: Sharp and clear\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Subdiaphragmatic Air/Abnormality: yes/no,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 11: \u0026apos;Bones and Soft Tissues:\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; Ribs: No fracture or lesion\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Spine: No abnormal curvature or compression\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Clavicles: Intact\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Shoulder Girdle: Unremarkable\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Soft Tissues: No swelling or subcutaneous air\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 12: \u0026apos;Supporting \u0026nbsp;Lines/Tubes/Devices:\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Central Line: Not present\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; ET Tube: Not present\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; NG Tube: Not present\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Pacemaker or Leads: Not present\\n\\\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Chest Tube: Not present\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;annotation_id2label = {\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 0: \u0026quot;Consolidation or Cavitory lesion\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 1: \u0026quot;Reticulonodular and ground glass opacity\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 2: \u0026quot;Cardiomegaly or chamber enlargement\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 3: \u0026quot;Pleural effusion\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 4: \u0026quot;Atelectasis\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 5: \u0026quot;Pleural thickening or Calcification\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 6: \u0026quot;Hilar lesions\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 7: \u0026quot;Fibrosis and fibrobronchiectatic changes\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 8: \u0026quot;Pneumothorax\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 9: \u0026quot;Mediastinal or Vascular and Aortic lesions\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 10: \u0026quot;Nodule\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 11: \u0026quot;Diaphragmatic or subdiaphragmatic abnormalities\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 12: \u0026quot;Mass\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 13: \u0026quot;Thoracic- cage or Soft tissue or Spinal abnormalities\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 14: \u0026quot;Opaque hemithorax\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 15: \u0026quot;Bronchial or Airway abnormalities\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 16: \u0026quot;Hernia\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 17: \u0026quot;Lung cyst or cystic lesions\u0026quot;,\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 18: \u0026quot;Supporting devices\u0026quot;}\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eHow we matched between the annotation ID with the report template\u0026nbsp;\u003c/p\u003e\n\u003cp\u003etemplate_02_id2annotation_id = {\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 1 : [15],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 2 : [6],\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp; \u0026nbsp; 3 : [\u003cem\u003e4,5,\u003c/em\u003e 7, 14, 17],\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 4 : [2, 9],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 5 : [0],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 6 : [1],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 7 : [10, 12],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 8 : [3],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 9 : [8],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u003cstrong\u003e10: [ 11, 14, 16],\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 11: [13],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 12: [18]}\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eannotation_id2template_02_id = {\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 0 : [5],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 1 : [6],\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 2 : [4],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 3 : [8],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 4 : [3],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 5 : [3],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 6 : [2],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 7 : [3],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 8 : [9],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 9 : [4],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 10: [7],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 11: [10],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 12: [7],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 13: [11],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 14: [3,10],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 15: [1],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 16: [10],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 17: [3],\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; 18: [12]}\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"TiruRad Diagnostics and research center","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Chest X-ray, Ethiopia, multicenter cohort, dataset, artificial intelligence, radiology","lastPublishedDoi":"10.21203/rs.3.rs-9417593/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9417593/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eChest radiography is widely used in Ethiopia for the evaluation of respiratory and cardiac disease. However, chest X-ray datasets used to develop and benchmark artificial intelligence (AI) systems are predominantly derived from high-income settings, which may limit generalizability and obscure region-specific radiographic patterns.\u003c/p\u003e\u003ch2\u003ePurpose\u003c/h2\u003e \u003cp\u003eTo describe multicenter patterns of chest X-ray imaging findings in Ethiopia using the Afro-Chest X-ray cohort and to summarize the radiologist-led workflow used to generate high-quality localization labels.\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003eThis retrospective multicenter study included deidentified chest X-rays acquired at 10 Ethiopian institutions from December 2022 through July 2025. After quality filtering, 55,409 chest X-rays were retained. Radiology reports were standardized using the Afro-Chest X-ray reporting template (Table\u0026nbsp;5 ). A stratified subset of 11,880 chest X-rays was manually annotated by 11 radiologists using bounding boxes for 19 thoracic findings (Table\u0026nbsp;3) with recorded confidence levels. Finding patterns was summarized descriptively using counts and proportions at the exam and finding-instance levels.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAmong 55,409 chest X-rays, 31,939 were linked to radiology reports from 48,962 patients (male, 18,324 [37.4%]; female, 30,387 [62.1%]). In the annotated subset (11,880 chest X-rays), 7,003 (58.9%) were abnormal and contained 22,531 labeled finding instances (mean, 3.22 instances per abnormal chest X-ray). The most frequent findings were reticulonodular or ground-glass opacities (n\u0026thinsp;=\u0026thinsp;6,896, 30.6%), pleural effusion (n\u0026thinsp;=\u0026thinsp;2,685, 11.9%), cardiomegaly or chamber enlargement (n\u0026thinsp;=\u0026thinsp;2,161, 9.6%), fibrosis or fibrobronchiectatic change (n\u0026thinsp;=\u0026thinsp;1,869, 8.3%), and consolidation with cavitary lesions (n\u0026thinsp;=\u0026thinsp;1,835, 8.1%).\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eIn this multicenter Ethiopian cohort, annotated abnormalities were dominated by parenchymal opacities, pleural effusion, and cardiomegaly. Afro-Chest X-ray provides a radiologist-verified reference for describing regional chest X-ray patterns and supporting the local validation and development of AI systems applicable to East and other sub-Saharan African clinical settings.\u003c/p\u003e","manuscriptTitle":"Multicenter Chest Radiographic Findings from Ethiopia with Radiologist-Generated Annotations","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-19 12:39:54","doi":"10.21203/rs.3.rs-9417593/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ea894265-ff2a-4d18-b84c-a4a1bb2dab91","owner":[],"postedDate":"April 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":66313050,"name":"Cardiac \u0026 Cardiovascular Systems"},{"id":66313051,"name":"Artificial Intelligence and Machine Learning"},{"id":66313052,"name":"Tropical Medicine"}],"tags":[],"updatedAt":"2026-04-19T12:39:54+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-19 12:39:54","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9417593","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9417593","identity":"rs-9417593","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}