Comparative Diagnostic Accuracy of Radiographic Techniques in Pediatric Chest Imaging

Thesis Abstract

The study addresses a critical gap in pediatric radiology by evaluating and comparing the diagnostic accuracy of multiple radiographic techniques used in pediatric chest imaging, with particular emphasis on balancing image quality, diagnostic yield, and radiation exposure. Pediatric chest radiography remains a cornerstone for initial assessment of respiratory illness, congenital anomalies, and trauma, yet variability in technique selection and patient-specific factors often results in inconsistent diagnostic performance. The aim is to determine which radiographic modalities—standard two-view radiography, lateral view, and high-resolution digital radiography (HRDR) with dose optimization—provide the highest diagnostic accuracy for common pediatric chest conditions, and to develop evidence-based guidance for modality selection. Specific objectives are (1) to quantify and compare sensitivity, specificity, positive and negative predictive values for each technique across a spectrum of pediatric chest pathologies including pneumonia, pneumothorax, pleural effusion, foreign body aspiration, and congenital defects; (2) to assess inter-observer agreement among pediatric radiologists and radiology residents; (3) to examine the role of patient factors (age, body habitus, cooperation level) on diagnostic performance; (4) to evaluate radiation dose implications associated with each technique; and (5) to propose an evidence-based decision model integrating diagnostic accuracy and radiation safety. The methodological approach adopts a cross-sectional diagnostic accuracy framework nested within a prospective multicenter study. The population comprises pediatric patients aged 0–18 years presenting with acute or chronic chest complaints at three tertiary-care centers over a 24-month enrollment period. A target sample size of 1,200 radiographs is planned to ensure precise estimates of sensitivity and specificity across subgroups, with stratified sampling to ensure representation by age groups and clinical indications. Data collection employs standardized imaging protocols for each technique, and radiographs are interpreted by a panel of five blinded readers, including two pediatric radiologists and three radiology residents. Reference standards combine chest computed tomography (CT) findings, clinical outcomes, microbiological data, and surgical/pathology reports where applicable to establish verifiable diagnostic truth. Data collection instruments include a structured reporting form capturing lesion characterization, confidence level, and reader recommendations. Validity and reliability are addressed through calibration sessions, intra- and inter-reader reliability assessment using kappa statistics, and periodic cross-checks against the reference standard. Analytical strategies encompass quantitative diagnostic accuracy metrics—sensitivity, specificity, likelihood ratios, diagnostic odds ratio, and area under the receiver operating characteristic (ROC) curve—for each technique and pathology. Comparisons across modalities employ McNemar tests for paired proportions and DeLong’s test for correlated ROC curves. Multivariate logistic regression models are constructed to adjust for confounders such as age, chest size, and cooperation level, while a mixed-effects model accounts for reader variability. A cost-utility angle is explored through a decision-analytic framework incorporating radiation dose metrics (volume CT dose index equivalents adapted for radiography), estimated lifetime cancer risk per modality, and resource utilization. The study integrates relevant theories, including signal detection theory to interpret reader performance and the ALARA principle within a risk-benefit analytic framework. Expected findings anticipate that HRDR with dose optimization will yield superior diagnostic accuracy for pneumonia and congenital anomalies with acceptable radiation exposure, while standard two-view radiography may suffice for straightforward cases such as pneumothorax detection in older children. Inter-observer agreement is hypothesized to be higher among experienced pediatric radiologists, with lower concordance among less experienced readers in subtle pathology. The results are anticipated to demonstrate a robust trade-off between diagnostic yield and radiation dose, informing a practical modality selection algorithm. The study contributes to knowledge by delivering empirically grounded, modality-specific performance benchmarks for pediatric chest imaging and by offering an evidence-based decision model to guide radiographic technique selection that optimizes diagnostic accuracy while minimizing radiation exposure. The findings are expected to influence clinical guidelines, radiology training curricula, and policy on pediatric imaging protocols, ultimately enhancing patient safety and diagnostic efficiency. Recommendations include adopting HRDR with dose optimization as the preferred initial modality for suspected pediatric chest pathology, while reserving standard radiography for low-risk presentations, accompanied by standardized reporting templates and reader calibration programs. Further research should explore integration with artificial intelligence-assisted image interpretation to augment diagnostic accuracy across modalities.