Artificial intelligence in radiology: a narrative review of current methods, clinical impact, and future directions

Abstract

This review synthesizes current artificial intelligence (AI) methodologies and evaluates their clinical impact in diagnostic radiology. As AI tools increasingly enter clinical workflows, understanding their performance, limitations, and barriers to adoption is critical. This review was conducted to provide a focused synthesis of recent advances in artificial intelligence (AI) as applied to diagnostic radiology. Relevant literature was identified through searches of PubMed, Scopus, and Google Scholar, covering publications from 2018 to 2025, using combinations of “artificial intelligence,” “machine learning,” “deep learning,” “radiology,” “medical imaging,” “workflow,” “ethics,” and “regulation.” Additional sources were located by screening reference lists of key papers and reviews. Articles were included if they explored clinical, technical, or ethical dimensions of AI within radiology, with emphasis on convolutional neural networks (CNNs), vision–language models (VLMs), workflow optimization, bias, or regulatory oversight (including FDA Software as a Medical Device [SaMD]). Only peer-reviewed, English-language articles with full-text availability were included. Editorials, commentaries, and studies outside clinical imaging were excluded. Rather than aiming for exhaustive coverage, the review integrates representative and influential work to illustrate current methods, clinical impact, and emerging challenges, with findings organized around diagnostic applications, operational efficiency, interpretability and bias, and ethics and regulation. Titles and abstracts were screened for relevance, and duplicates were removed. Eligible studies were synthesized thematically rather than quantitatively to highlight clinical applicability and reproducibility. Rather than aiming for exhaustive coverage, the goal of this review was to integrate the most relevant and representative work that illustrates current methods, clinical impact, and emerging challenges. The literature was synthesized narratively, with findings organized around four central themes: diagnostic applications, operational efficiency, interpretability and bias, and ethics and regulation. This approach follows the SANRA (Scale for the Assessment of Narrative Review Articles) guidelines to ensure transparency and scholarly rigor. AI improves diagnostic sensitivity, prioritizes critical cases, and optimizes radiology workflows. Clinical uses span triage, image classification, report generation, and administrative task automation. However, challenges persist, including limited model interpretability, lack of external validation, demographic bias, and inconsistent regulatory frameworks. Vision-language models expand AI capabilities but show reduced accuracy in underrepresented populations and lack contextual clinical reasoning. Representative studies supporting these findings are summarized in Table 2, with reported AUC values for key diagnostic models ranging from 0.84 to 0.96 across tasks such as hemorrhage detection, mammography, and chest X-ray triage. AI is transforming radiology into a hybrid of machine-driven augmentation and clinical oversight. Successful implementation requires not only algorithmic accuracy but also interpretability, fairness, and ethical infrastructure. Radiologists must play a central role in validating and supervising these technologies to ensure they support patient-centered, equitable care. Artificial intelligence is rapidly transforming radiology workflows, and this review clarifies how clinicians can guide its ethical and effective clinical integration. Not applicable. • AI tools are actively reshaping diagnostic radiology through triage, detection, and workflow optimization. • Convolutional and vision-language models improve sensitivity but raise concerns about generalizability and fairness. • Radiologists must supervise AI implementation to preserve accountability and ensure patient-centered care. • Administrative AI improves efficiency in scheduling, worklist triage, and reporting. • Ethical oversight and transparency are essential for safe and equitable AI deployment in medical imaging.

Ähnliche Arbeiten