Photodynamic therapy (PDT) has emerged as a promising non-invasive therapeutic modality for cancer treatment. Compared with conventional antineoplastic therapies, PDT offers targeted cytotoxic effects and relatively low systemic toxicity, making it an attractive therapeutic option for patients with cancer. The therapeutic efficacy of PDT is largely determined by the intrinsic properties of photosensitizers (PSs), which are the central components of this therapeutic approach. However, currently available PSs still present several inherent limitations, including inadequate penetration into deep tumor tissues, suboptimal accumulation and delivery at tumor sites, and substantial dependence on oxygen availability within the tumor microenvironment. Consequently, the development of high-performance PSs remains a key research priority in the field of PDT. Recent advances in photosensitizer technology have enabled the rational integration of PDT with other therapeutic modalities, including radiotherapy, chemotherapy, and immunotherapy, to achieve synergistic therapeutic effects, thereby significantly improving overall treatment outcomes. This review systematically summarizes the major classifications, key characteristics, and recent advances in the application of nanomaterials for PDT. It evaluates the current progress in the development of nanomaterial-based PSs and examines their potential applications in combination therapeutic strategies for cancer treatment. Collectively, these findings provide valuable insights into the future development of PDT-based anticancer therapies. Photodynamic therapy (PDT) is a minimally invasive and targeted therapeutic approach for cancer treatment.Nanomaterial-based photosensitizers enhance tumor targeting, bioavailability, and photodynamic efficacy.PDT exerts antitumor effects through direct cytotoxicity, vascular disruption, and immune activation.Combination strategies integrating PDT with chemotherapy, radiotherapy, and immunotherapy demonstrate synergistic antitumor efficacy.
Nearly 100 years from its birth, radiology continues to grow as though still in adolescence. Although some radiologic technologies have matured more than others, new applications and techniques appear regularly in the literature. Radiology has evolved from purely diagnostic devices to interventional technologies. New contrast agents in MRI, X ray and ultrasound enable physicians to make diagnoses and plan therapies with greater precision than ever before. Techniques are less and less invasive. Advances in computer technology have given supercomputer-like power to high-end nuclear medicine and MRI systems. Imaging systems in most modalities are now designed with upgrades in mind instead of "planned obsolescence." Companies routinely upgrade software and other facets of their products, sometimes at no additional charge to existing customers. Hospitals, radiology groups and imaging centers will face increasing demands to justify what they do according to patient outcomes and management criteria. Did images make the diagnosis or confirm it? Did the images determine optimal treatment strategies or confirm which strategies might be appropriate? Third-party payers, especially the government, will view radiology in those terms. The diagnostic imaging and therapy systems of today require increasingly sophisticated technical support for maintenance and repair. Hospitals, radiology groups and imaging centers will have to determine the most economic and effective ways to guarantee equipment up-time. Borrowing from the automotive industry, some radiology manufacturers have devised transtelephonic software systems to facilitate remote troubleshooting. To ensure their fiscal viability, hospitals continue to acquire new imaging and therapy technologies for competitive and access-to-services reasons.(ABSTRACT TRUNCATED AT 250 WORDS)
Breast cancer screening with mammography reduces mortality, but interpretation errors persist. While artificial intelligence (AI) shows promise, most studies focus on case-level accuracy and overlook lesion-level performance. This study evaluated the AI (GMIC+CLAHE) model against radiologists and radiology trainees, considering breast density and lesion characteristics. Nine BREAST test sets of screening digital mammograms (540 cases:179 cancer, 361 normal) were analysed. AI performance was compared with 17 radiologists and 26 trainees. Case- and lesion-level performance were assessed across breast density, lesion type, and size. AI malignancy scores were dichotomised using Youden's index. Sensitivity, specificity, ROC AUC, lesion localisation, and odds ratios (ORs) were calculated. AI achieved higher case-level AUC (0.974) than radiologists (0.82±0.06;P=0.01) and trainees (0.74±0.05;P<0.0001) across breast densities. At lesion level, AI matched radiologists and outperformed trainees (OR = 1.6;P=0.03), with lesion sensitivities of 66.3% (AI), 67.0% (radiologists), and 55.0% (trainees). AI also outperformed trainees in detecting mixed-type lesions (75%-vs-55%;OR=2.4) and small lesions ≤15 mm (65.7%-vs-55%;OR=1.6). Architectural distortion had the highest miss rate (43.8%) whereas mixed-type lesions demonstrated the lowest miss rate (25.0%) by AI. AI demonstrated superior case-level accuracy compared with radiologists and trainees, but limitations in localising specific lesion types, supporting its role as an adjunct to human readers in screening and training. This study integrates case-and-lesion-level evaluation in different case characteristics, identifying key strengths and limitations for clinical implementation. Despite superior case-level accuracy, variability in lesion localisation across cancer types and sizes highlights important considerations for its integration into screening services.
Dedifferentiated chondrosarcoma (DDCS) remains one of the most aggressive and therapeutically challenging primary bone malignancies. While surgery continues to represent the cornerstone of treatment for localized disease, outcomes remain poor due to high rates of local recurrence and metastatic progression. The role of adjuvant therapies remains less clear. Radiotherapy may be considered in selected patients with unresectable tumors, positive margins, or for symptom palliation, whereas conventional chemotherapy has demonstrated inconsistent and generally limited benefit despite the use of osteosarcoma-based regimens. Recent advances in molecular profiling have improved our understanding of DDCS biology and revealed potential therapeutic opportunities. Recurrent IDH1/2 mutations appear to contribute to tumor progression and dedifferentiation and may represent actionable therapeutic targets. In our practice, comprehensive molecular profiling should be considered whenever feasible, particularly in advanced disease. Emerging evidence suggests that immunotherapy can induce durable responses in a small subset of patients, while early studies combining immune checkpoint inhibitors with antiangiogenic tyrosine kinase inhibitors have demonstrated encouraging response rates that exceed those historically achieved with chemotherapy alone. Although these findings require prospective validation, they support a shift toward biomarker-driven and combination treatment strategies.
Lung cancer is the most common cause of cancer death in developed countries. The prognosis is poor, with less than 15% of patients surviving 5 years after diagnosis. The poor prognosis is attributable to lack of efficient diagnostic methods for early detection and lack of successful treatment for metastatic disease. Most patients (>75%) present with stage III or IV disease and are rarely curable with current therapies. Within the last decade, rapid advances in molecular biology, pathology, bronchology, and radiology have provided a rational basis for improving outcome. These advancements have led to a better documentation of morphological changes in the bronchial epithelium before development of clinical evident invasive carcinomas. This has changed our concept of lung carcinogenesis and emphasized the multistep carcinogenesis approach on several levels. Combined with the technical developments in bronchoscopic techniques, e.g., laser-induced fluorescence endoscope (LIFE) bronchoscopy, we now have improved methods to localize preinvasive and early-invasive bronchial lesions. With the LIFE bronchoscope, a new morphological entity (angiogenic squamous dysplasia) has been recognized, which might be an important biomarker and target for antiangiogenic chemopreventive agents. To reduce the mortality of lung cancer, these new technologies have been taken into the clinic in different scientific settings. The use of low-dose spiral computed tomography in the screening of a high-risk population has demonstrated the possibility of diagnosing small peripheral tumors that are not seen on conventional X-ray. A shift in the therapeutic paradigm from targeting advanced clinically manifest lung cancer toward asymptomatic preinvasive and early-invasive cancer is occurring. The present article reviews the recent advances in the diagnosis of preinvasive and early-invasive cancer to identify biomarkers for early detection of lung cancer and for chemoprevention studies.
In the clinical management of prostate cancer, continuously refined scoring systems form the cornerstone of precise risk stratification. Pathological assessment systems, represented by the Gleason score and the International Society of Urological Pathology (ISUP) grade groups, have long been the cornerstone of prostate cancer risk stratification; however, they are subject to inherent limitations such as sampling bias and spatial heterogeneity. With advances in imaging technology, the PI-RADS scoring system based on multiparametric magnetic resonance imaging (mpMRI) has been widely adopted for the detection and localization of prostate cancer; however, inter-observer variability remains a significant issue in clinical practice. Concurrently, emerging functional imaging scoring systems based on PSMA-PET, such as PSMA-RADS and miTNM staging, have provided new tools for the precise staging of prostate cancer and restaging following biochemical recurrence. Currently, an increasing number of studies are attempting to integrate pathological and multimodal imaging information to construct comprehensive predictive models, thereby addressing the shortcomings of single-modality assessment methods, achieving more personalized and precise risk stratification, and ultimately optimizing treatment decisions and improving patient prognosis.
Systemic vasculitides are complex multisystem immune-mediated inflammatory diseases, classified by the affected vessel size, which are associated with a wide range of clinical manifestations and complications. The most common forms of large vessel vasculitis (LVV) are giant cell arteritis and Takayasu arteritis, although other more rare causes exist. Cranial giant cell arteritis is the most commonly diagnosed form of LVV, affects older adults and can result in sudden visual loss and stroke. Up to 80% of individuals with giant cell arteritis have LV involvement, which can exist with or without cranial involvement. Takayasu arteritis usually occurs in those under the age of 60 years, resulting in progressive injury of the aorta and its main branches. LVV may also affect the coronary and pulmonary circulations, as well as the pericardium and myocardium resulting in a range of cardiovascular pathologies. Diagnosing LVV can be challenging in the absence of a disease-specific laboratory biomarker, and relies on a combination of inflammatory markers, imaging, and temporal artery biopsy when required. Noninvasive multimodality imaging techniques are advancing and provide new avenues for diagnosis and disease monitoring. High-resolution ultrasound, magnetic resonance imaging, and computed tomography can survey the pattern and extent of arterial involvement and reveal signs of active inflammation. Positron emission tomography imaging with 18F-fluorodeoxyglucose and novel radiotracers provide more sensitive measures of vascular inflammation of the large vessels. Ultimately, these imaging tests can be used to guide therapeutic interventions and inform the clinical use of new targeted disease modifying therapies for LVV.
Oxygen extraction fraction (OEF) is a physiological parameter reflecting the fraction of delivered arterial oxygen extracted by cerebral tissue, providing information complementary to perfusion, structural MRI, and conventional BOLD contrast. Positron emission tomography (PET) remains the reference standard for quantitative assessment of OEF and cerebral metabolic rate of oxygen (CMRO₂), but limited availability, radiotracer requirements, and operational complexity have motivated MRI-based alternatives. This review focuses on MRI-derived OEF mapping using quantitative BOLD (qBOLD), quantitative susceptibility mapping (QSM), hybrid QSM-qBOLD (QQ) methods, constrained qBOLD, Bayesian/prior-based reconstruction, and artificial-intelligence-assisted approaches. Classical qBOLD provides a model-based route for estimating oxygenation-related parameters from deoxyhemoglobin-induced signal decay, but remains limited by parameter degeneracy, noise sensitivity, and assumptions regarding venous blood volume, hematocrit, relaxation, and vascular geometry. QSM provides complementary susceptibility information from gradient-echo phase data; however, total susceptibility cannot independently distinguish heme-related venous deoxygenation from non-heme sources such as iron, myelin, calcification, or tissue composition. Hybrid QSM-qBOLD methods address this limitation by jointly using magnitude and susceptibility information to improve susceptibility-source separation and reduce unrealistic assumptions. Recent developments, including constrained qBOLD, temporal clustering, multi-echo complex QQ, Bayesian reconstruction, ANN-based inference, and QQ-NET, aim to improve stability, computational efficiency, and clinical practicality. Validation remains a central challenge because direct MRI-PET comparison studies are limited. Clinical applications are most promising in disorders where oxygen delivery, extraction, and metabolism become uncoupled.
Cardiac magnetic resonance (CMR) imaging has become one of the most comprehensive noninvasive tools for evaluating cardiac structure, function, and myocardial tissue characteristics. Moreover, the role of CMR imaging in cardiovascular tomography has become increasingly central, providing unparalleled insights across a wide range of cardiac diseases. Over the past decade, several technical advances have expanded the clinical applications of CMR imaging and strengthened the role of this modality in diagnosis and management. Therefore, this review synthesizes these rapid advances and the associated growing influence on contemporary cardiovascular imaging practice. This literature review examines recent advances in CMR imaging, including improved contrast agents, refined mapping techniques, integration of artificial intelligence (AI), and accelerated imaging protocols, and highlights how these innovations are enhancing CMR imaging performance and delivering meaningful benefits in clinical practice. Recent advances in CMR imaging include higher-resolution imaging, novel gadolinium-enhanced and parametric mapping sequences, strain imaging, and four-dimensional (4D) flow magnetic resonance imaging (MRI), enabling more comprehensive assessment of myocardial structure, function, and hemodynamics. Improvements in contrast agents, the development of gadolinium alternatives, and increasing integration of AI have further enhanced the diagnostic accuracy and workflow efficiency. Advances in CMR imaging have strengthened the role of this technique as a highly accurate and increasingly accessible tool for cardiovascular evaluation, despite ongoing challenges related to cost, patient tolerance, and contrast use. Future directions should include enhanced tissue characterization, integration with complementary imaging modalities, and continued technological innovation to expand clinical applications, improve patient outcomes, and reduce reliance on radiation-based techniques.
More than 300 million computed tomography (CT) scans are performed worldwide each year, yet many early or incidental tumors in these scans remain undetected. Artificial intelligence (AI) could help: segmentation models can surpass radiologists and alternative AI models in detecting tumors, and they localize the tumors for radiologist verification. However, segmentation-based tumor detection has long been limited by the need for tumor masks: radiologist-drawn tumor outlines that are scarce, expensive, and entirely unavailable for many cancer types. In contrast, nearly every CT scan is accompanied by a radiology report with detailed tumor descriptions. Yet, these reports have not been used effectively to train tumor segmentation models. Here, we introduce R-Super, a framework that converts routine radiology and pathology reports into localized training signals for tumor segmentation. R-Super trains AI to segment tumors that match their descriptions in reports. Reports are only needed for training, not inference. We trained R-Super on 127,496 CT-Report pairs (42 million 2D images, USA) and evaluated it internally at UCSF ( N = 2,301 , USA) and externally at Stanford ( N = 1,976 , USA), Medipol ( N = 1,327 , Turkey), and Basel ( N = 2,935 , Switzerland). R-Super detects 7 tumor types for which public tumor masks are scarce or absent: spleen, gallbladder, prostate, bladder, uterus, esophagus, and adrenal tumors. Training R-Super on over 100,000 reports (no mask) outperformed mask-based segmentation models trained on 870 masks, demonstrating that large-scale report-based training surpasses smaller-scale mask-based training. Alternatively, by training R-Super on both these reports and masks together, cancer detection sensitivity increased by over +11% beyond mask-only training, and DSC by +14%. R-Super significantly surpassed 6 alternative training frameworks trained on the same dataset and 9 leading public AI models. R-Super significantly surpassed six radiologists in detecting six tumor types and matched them for uterus tumors in a reader study. On average, R-Super detected 56% more malignant tumors than the radiologists, at matched false positive rate. These results show that radiology reports are not merely clinical documentation, but a large-scale, underutilized source of localized supervision for training more accurate cancer detection AI. By effectively learning from reports, R-Super enables tumor segmentation models to scale beyond scarce radiologist-drawn tumor masks and advances automated and incidental cancer detection closer to clinical deployment. We release code, over 22,000 CT scans and reports, and the first public AI model to reach or surpass radiologist performance in detecting these tumor types on CT.
Advances in data science and medical artificial intelligence (AI) raise complex philosophical and ethical quandaries about what it means to know a person or a community through data and what kinds of people and societies we are becoming in this era of predictive data science. Drawing on four lightly fictional but reality-informed case studies in mental health, radiology, genomics and environmental public health, we reflect on how AI technologies, largely built on Western biomedical traditions, may conflict with relational, spiritual and Indigenous understandings of health and wellbeing. This may manifest as epistemic friction, algorithmic fatalism and diminished trust in patient-clinician relationships. Besides familiar concerns regarding bias and transparency, this paper advances the discourse on the ethics of medical AI and data science in healthcare by shifting the analysis from epistemology (how AI systems know, classify and predict) to ontology (the study of the nature of being, as reconfigured by data and AI). We argue that AI systems may inflict significant ontological harm by reconfiguring identity, moral agency and imagined futures and advocate for a renewed medical humanism driven by inter- philosophies dialogue, cross-cultural ethics and ecocentric approaches to care. From Bamenda, Cameroon, to Mthatha, South Africa and Toronto, Canada, the future of medical AI must be defined by the moral and philosophical traditions that people already live by. African, Indigenous, Islamic, Buddhist, Confucian and marginalised Western worldviews should not be treated as peripheral critiques, but as constitutive resources for building inclusive, human-centred health technologies. We conclude that bioethics should be recognised as a core infrastructure in global health, on equal footing with data science, medicine, biomedical research and health innovation.
Artificial intelligence (AI) has become firmly established in radiology, with most current applications relying on task-specific convolutional neural networks (CNNs) such as U-Net architectures for segmentation, detection, and classification. The emergence of foundation models (FMs), however, marks a fundamental paradigm shift. These large-scale, pretrained models are designed to flexibly adapt to a wide range of radiological tasks, often requiring only minimal task-specific fine-tuning. In medical imaging, FMs are typically trained on large, heterogeneous, and multimodal data sets and enable novel applications such as zero-shot segmentation, prompt-based image exploration, and the integration of clinical context information. In parallel with academic developments, the first vendors have begun to incorporate FM-based approaches in commercial radiology products, gradually replacing narrowly specialized model architectures. For clinical practice, this technological shift offers potential benefits, including reduced false-positive findings, automated triage in emergency imaging, and improved detection of complex or atypical pathologies. At the same time, these advances come with increased demands on computational resources and costs, as well as unresolved challenges regarding validation, explainability, and regulatory approval. This review introduces the core principles of foundation models, discusses their technical and economic implications, and illustrates - using clinically relevant examples - how this new generation of models may fundamentally transform radiological image analysis. · Unlike conventional task-specific networks such as U-Net, foundation models can be flexibly applied and adapted to different radiological tasks.. · Foundation models may improve radiological and clinical performance, particularly in complex or data-limited settings.. · Radiologists and industry partners must critically validate foundation-model outputs and ensure their explainable integration into clinical workflows in accordance with regulatory requirements.. · Shahzadi I, Borggrefe J. Fundamentals of Foundation Models in Radiological Image Processing: Opportunities, Limitations, and Clinical Case Studies. Rofo 2026; DOI 10.1055/a-2899-1290. Künstliche Intelligenz (KI) ist in der Radiologie mittlerweile fest etabliert, wobei bislang vor allem aufgabenspezifische Convolutional Neural Networks (CNNs) wie U-Net-Architekturen für Segmentierung, Detektion und Klassifikation eingesetzt werden. Mit dem Aufkommen sogenannter Foundation Models (FMs) vollzieht sich jedoch ein grundlegender Paradigmenwechsel: Diese großskaligen, vortrainierten Modelle sind darauf ausgelegt, sich flexibel an eine Vielzahl radiologischer Aufgaben anzupassen, häufig mit nur minimalem aufgabenspezifischem Fine-Tuning. In der medizinischen Bildgebung basieren FMs typischerweise auf heterogenen, multimodalen Datensätzen und ermöglichen neue Anwendungen wie Zero-Shot-Segmentierung, prompt-basierte Bildexploration und die Integration klinischer Kontextinformationen. Parallel zur akademischen Entwicklung beginnen erste Hersteller, FM-basierte Ansätze in kommerzielle radiologische Produkte zu integrieren und klassische, eng spezialisierte Modellarchitekturen schrittweise abzulösen. Für die klinische Praxis ergeben sich daraus potenzielle Vorteile, etwa in der Reduktion falsch-positiver Befunde, der Automatisierung der Triage in der Notfalldiagnostik oder der verbesserten Detektion komplexer Pathologien. Demgegenüber stehen jedoch erhöhte Anforderungen an Rechenressourcen, Kosten sowie bislang ungelöste Fragen der Validierung, Erklärbarkeit und regulatorischen Einordnung. Diese Übersichtsarbeit stellt die konzeptionellen Grundlagen von Foundation Models vor, diskutiert deren technische und ökonomische Implikationen und illustriert anhand klinisch relevanter Anwendungsbeispiele, wie diese neue Modellgeneration die radiologische Bildanalyse nachhaltig verändern könnte. · Im Gegensatz zu konventionellen aufgabenspezifischen Netzwerken wie U-Net können Foundation Models flexibel für verschiedene radiologische Aufgaben eingesetzt und angepasst werden.. · Foundation Models können die radiologische und klinische Leistungsfähigkeit insbesondere bei komplexen oder datenarmen Fragestellungen verbessern.. · Radiologinnen und Radiologen sowie Industriepartner müssen die Ergebnisse von Foundation Models kritisch validieren und ihre nachvollziehbare Integration in klinische Arbeitsabläufe unter Einhaltung regulatorischer Anforderungen sicherstellen.. · Shahzadi I, Borggrefe J. Fundamentals of Foundation Models in Radiological Image Processing: Opportunities, Limitations, and Clinical Case Studies. Rofo 2026; DOI 10.1055/a-2899-1290.
Surgical expertise has traditionally rested on cumulative clinical knowledge and technical experience. Advances in artificial intelligence, robotics, computer vision, intraoperative analytics, and sensor-rich operating environments may reshape the foundational characteristics of surgical performance. A narrative review of emerging surgical technologies was performed alongside historical parallels from aviation, manufacturing, diagnostic radiology, geospatial sciences, and military decision systems with resultant identification of four themes characterizing how expertise may migrate from manual execution toward cognitive integration and system stewardship. Across prior transitions, mature technologies reduced manual variability, elevated supervisory competencies, generated new perceptual paradigms, and enabled distributed expertise. If these patterns apply to surgery, they suggest a potential shift toward decision-making under uncertainty, anomaly management, system orchestration, and multimodal data integration. As intelligent systems assume greater operative roles, defining surgical competencies may increasingly become cognitive, integrative, and systems-level. Training, credentialing, workforce models, and policy should anticipate this potential evolution to ensure human-machine partnerships enhance safety, equity, and value.
The availability of evidence-informed frameworks to support learning interventions in dental radiography for diagnostic radiographers is limited. The framework used in this study was guided by participatory action research and adapted using the precede-proceed model. The framework is underpinned by social justice and systems and learning theory. This study aimed to determine and describe the application of a framework for supporting dental radiography as part of diagnostic radiography practice. Public and private radiology departments selected higher educational institutions from provinces of KwaZulu-Natal, Gauteng and Western Cape, inclusive of members from the Radiography and Clinical Technology Board at the Health Professions Council of South Africa. A sequential mixed-methods design was used for framework development, guided by the Precede-Proceed Model. Stage 1 involved a needs assessment, which comprised a quantitative Phase 1 and a qualitative Phase 2, which included an online questionnaire and semi-structured interviews, respectively. Stage 2 focused on the planning and delivery of the online Continuous Professional Development programme. Stage 3 involved a post-programme quantitative questionnaire to evaluate participant feedback and learning outcomes. A strength of the framework is professional resilience and adaptability, enabling adaptation to technological advances and the changing needs of patients. The framework proposes a practical way to form targeted continuous professional development (CPD) initiatives, making it valuable in resource-constrained environments. The framework is beneficial for supporting dental radiography training and practices and promoting equitable access to dental radiography. This framework adds value for supporting the practice of dental radiography for the diagnostic radiographers.
Primary liver cancers, including hepatocellular carcinoma and intrahepatic cholangiocarcinoma, are one of the leading causes of cancer-related mortality. Transarterial chemoembolisation (TACE) and radioembolisation (TARE) are established palliative treatments yet they are traditionally performed during inpatient hospitalisation. Recent technical and organisational advances allow for safe same-day ambulatory procedures. In particular, the extent to which ambulatory intra-arterial therapy aligns with patients' expectations, comfort and quality of life remains poorly documented. The Care pathway for Hepatic intra-arterial Oncology in an ambulatory Context study (CHOC) trial aims to evaluate the implementation and effectiveness from both patient-centred and clinical perspectives of an ambulatory care pathway for intra-arterial treatment of primary liver cancer in a multicentre randomised hybrid type 1 trial. CHOC is a pragmatic, multicentre, randomised controlled hybrid type 1 trial comparing ambulatory versus conventional inpatient care for patients undergoing TACE or TARE for primary liver cancer. A total of 206 patients (103 per arm) will be randomised 1:1 and followed for 7 months. The primary outcome is the patient's global satisfaction score, measured 3 days post-procedure using the EORTC PATSAT-C33 questionnaire. Secondary outcomes include quality of life, safety, clinical outcomes, cost analysis and an embedded qualitative implementation study assessing acceptability, adoption, feasibility and sustainability across centres. Economic analyses will estimate both per-patient costs and the 5-year budget impact for the French national health insurance. The study has received ethical approval from the Protection of Persons Committee and adheres to the Declaration of Helsinki, good clinical practice and French regulatory requirements. Findings will be disseminated through peer-reviewed publications, conferences and participating centres to guide broader implementation of ambulatory interventional radiology care. NCT06990659.
The increasing integration of Artificial Intelligence (AI) into clinical workflows for medical imaging and radiotherapy presents new opportunities and challenges for the radiation protection of patients, staff, and the public. This perspective from the International Commission on Radiological Protection (ICRP) Committee 3 Working Party on AI examines how current and emerging AI applications support the core principles of justification and optimisation across diagnostic and interventional radiology, nuclear medicine, and radiotherapy, and identifies priorities for their safe clinical implementation. AI applications with the greatest current clinical maturity include clinical decision support for referral appropriateness, image reconstruction, protocol optimisation, automated contouring, treatment planning, adaptive radiotherapy, and AI-enabled quality assurance. Other applications, including patient-specific dosimetry, occupational dose prediction, synthetic imaging, and predictive safety analytics, show considerable promise but remain at earlier stages of validation. Significant challenges accompany these advances: data biases and limited generalisability may undermine performance across diverse settings; the "black box" nature of many models complicates clinical accountability; and robust validation, continuous quality assurance, and harmonised regulatory oversight remain essential. Dedicated training in AI literacy for healthcare professionals is critical for safe deployment. AI has great potential to improve medical radiation protection but its safe use requires careful management of associated risks. By identifying areas of established clinical adoption, emerging applications, and common implementation priorities, this perspective provides a framework to support future ICRP recommendations on the safe integration of AI into medical radiation protection.
The American College of Radiology BI-RADS v2025 Manual, representing the first major BI-RADS update since the 5th edition from 2013, incorporates advances in breast imaging technology, clinical practice, and regulatory requirements. This AJR Expert Panel Narrative Review describes the manual's most clinically significant updates for each modality, highlighting implications for interpretation, reporting, and patient management in everyday practice. Significant revisions were introduced across the mammography, ultrasound, MRI, contrast-enhanced mammography (CEM), and auditing sections to improve standardization and reporting consistency. Key mammography updates include alignment of breast density terminology with Mammography Quality Standards Act regulations, expanded digital breast tomosynthesis guidance, and removal of the standalone descriptor "developing asymmetry." Ultrasound updates include incorporation of nonmass lesions, elasticity descriptors, and automated breast ultrasound terminology. MRI updates include revisions to the terminology for enhancing lesions and expanded implant descriptors. BI-RADS v2025 also formally incorporates CEM as a dedicated modality section. Lexicon terminology and reporting principles have been harmonized across modalities. Additional revisions to auditing and outcomes monitoring emphasize modality-specific performance metrics and quality assurance. The BI-RADS v2025 updates are expected to enhance communication, reporting reproducibility, and quality assessment in current breast imaging practice while providing a flexible framework for future modifications and enhancements.
Substantial recent advances in percutaneous image-guided minimally invasive interventions provide a robust arsenal for radiologists in management of patients with bone tumors. Percutaneous minimally invasive thermal ablation, cementation with or without osseous reinforcement via implants, and osteosynthesis have been successfully used and progressively incorporated into the management paradigm of patients with bone tumors. The purpose of this article is to outline the recent advances in such interventions and the role of radiologists in managing patients with bone tumors in modern-era practice.
Autosomal recessive polycystic kidney disease (ARPKD) is a rare, genetic ciliopathy with significant associated morbidity and mortality. ARPKD kidney disease is characterized by diffuse kidney fusiform collecting duct dilatations (microcysts) leading to progressive kidney dysfunction. Patients with ARPKD can also develop clinically-significant congenital hepatic fibrosis and portal hypertension. Despite advances in clinical management, there are currently no disease-modifying therapies approved for ARPKD, and the development of clinical trials has been limited by the lack of sensitive, non-invasive biomarkers capable of detecting early disease progression in both the kidney and liver. MRI provides a unique opportunity to address this gap by offering multiparametric, quantitative assessments of tissue structure and function without ionizing radiation or, in many cases, contrast agents. This review summarizes emerging quantitative MRI-based techniques that show promise as imaging biomarkers for children and young adults with ARPKD. These advanced MRI approaches enable comprehensive characterization of the multi-faceted structural and functional changes in the kidney and liver associated with ARPKD. Continued development and harmonization of these quantitative imaging biomarkers across institutions will be critical for enabling multi-center studies and supporting future therapeutic trials aimed at improving outcomes for children and young adults with ARPKD.
Liver biopsy remains essential for diagnosing and managing pediatric liver diseases, despite advances in noninvasive techniques. However, data on the current indications, diagnostic yield, and safety of ultrasound-assisted percutaneous liver biopsy in children are limited. We retrospectively analyzed 60 pediatric patients who underwent ultrasound-assisted percutaneous liver biopsy at Agia Sofia Children's Hospital between January 2018 and September 2025. Data collected included demographics, laboratory values, biopsy indications, histopathology and complications. Biopsies were performed under anesthesia using an 18-G Tru-Cut needle with real-time ultrasound guidance. Hemoglobin and hematocrit were measured pre-procedure and at 4 and 24 h post-procedure. Specimen adequacy was assessed by core length and number of complete portal tracts (CPTs). The mean patient age was 8.6 years (range 1 month to 17 years), with weights from 2.9-65 kg. Indications included persistent hypertransaminasemia, cholestasis, autoimmune and metabolic liver disease, and post-transplant evaluation. Mean biopsy length was 1.12 cm (range 0.5-2 cm), with a mean of 11 CPTs (range 4-17). Histopathology revealed autoimmune hepatitis (AIH) in 38.3%, AIH with primary sclerosing cholangitis in 13.3% of patients with inflammatory bowel disease, and other diagnoses in the remainder. No major complications or postprocedural bleeding occurred. Hemoglobin and hematocrit remained stable, and all specimens were adequate for histologic assessment. Ultrasound-assisted percutaneous liver biopsy with an 18-G needle is safe, and yields sufficient tissue for diagnosis in pediatric patients. It continues to be indispensable for the definitive diagnosis, staging, and management of diverse liver diseases in children.