Crossing the Ai Innovation-To-Application Chasm Through Implementation Science: Using Concept Mapping in the Knowledge-To-Action Process to Identify and Overcome Barriers to Routine Use of Effective Ai in Medical Imaging Publisher



Jouzdani AF ; Gorji A ; Zahedi M ; Alizadeh M ; Saboury B ; Rahmim A ; Fayazbakhsh A ; Salmanpour MR
Source: Progress in Biomedical Optics and Imaging - Proceedings of SPIE Published:2026

Abstract

Timely and accurate interpretation of medical imaging studies is central to high-quality patient care, particularly in oncology, neurology, and cardiology, where imaging drives life-altering diagnostic and treatment decisions. Modern imaging modalities such as PET, SPECT, CT, MRI, ultrasound, and X-ray generate volumes of complex data that demand rapid and precise interpretation to prevent delays, reduce errors, and optimize workflows. Radiologists and nuclear medicine physicians face pressures from rising workloads, workforce shortages, and the cognitive burden of high-stakes interpretation. These challenges create opportunities for Artificial Intelligence-based Diagnostic Decision Support (AI-DDS) tools, which can enhance accuracy, accelerate detection of subtle findings, standardize reporting, and bridge the gap between imaging data and actionable insights. By complementing expert judgment with automated analytics, AI-DDS can reduce variability, improve efficiency, and advance patient outcomes. Despite this promise, AI-DDS adoption in medical imaging remains slow, hindered by infrastructure limitations, workflow integration challenges, insufficient training, lack of trust, and ethical or legal concerns, collectively forming the “AI chasm” in radiology and nuclear medicine. To systematically identify and prioritize these barriers, we employed implementation-science-guided concept mapping (CM) within the Knowledge-to-Action (KTA) framework, a structured methodology for translating innovation into clinical practice. Through a systematic literature review and interviews with 12 imaging experts, including five nuclear medicine specialists, we generated 96 validated statements capturing the multifaceted challenges of AI integration. These statements were synthesized, conceptually grouped, and analyzed using multidimensional scaling and hierarchical clustering, revealing 11 barrier clusters spanning infrastructure gaps, modality-specific adaptability limitations, ethical and legal uncertainties, insufficient AI literacy, and professional resistance. By embedding CM within the KTA framework, this study delivers a comprehensive, modality-focused roadmap for integrating AI-DDS into clinical imaging. Identifying these 11 barrier clusters provides actionable guidance for institutions, policymakers, and stakeholders to overcome adoption hurdles, accelerate explainable AI deployment, and improve patient care. © 2026 SPIE. All rights reserved.