Canadian Association of Radiologists Journal-Journal De L Association Canadienne Des Radiologistes最新文献

Purpose: Large language models (LLMs) have the potential to support clinical decision-making but often lack training on the latest clinical guidelines. Retrieval-augmented generation (RAG) may enhance guideline adherence by dynamically integrating external information. This study evaluates the performance of two LLMs, GPT-4o and o1-mini, with and without RAG, in adhering to Canadian radiology guidelines for incidental hepatobiliary findings. Methods: A customized RAG architecture was developed to integrate guideline-based recommendations into LLM prompts. Clinical cases were curated and used to prompt models with and without RAG. Primary analyses assessed the rate of guideline adherence with comparisons made between LLMs with and without RAG. Secondary analyses evaluated reading ease, grade level, and response times for generated outputs. Results: A total of 319 clinical cases were evaluated. Adherence rates were 81.7% for GPT-4o without RAG, 97.2% for GPT-4o with RAG, 79.3% for o1-mini without RAG, and 95.1% for o1-mini with RAG. Model performance differed significantly across groups, with RAG-enabled configurations outperforming their non-RAG counterparts (P < .05). RAG-enabled models demonstrated improved reading ease and lower grade level scores; however, all model outputs remained at advanced comprehension levels. Response times for RAG-enabled models increased slightly due to additional retrieval processing but remained clinically acceptable. Conclusions: RAG-enabled LLMs significantly improved adherence to Canadian radiology guidelines for incidental hepatobiliary findings without compromising readability or response times. This approach holds promise for advancing evidence-based care and warrants further validation across broader clinical settings.

Purpose: To evaluate frequency and distribution of immune-related adverse events detected by ¹⁸F-FDG PET/CT in patients with metastatic melanoma undergoing immunotherapy. Materials and Methods: Retrospective observational cohort study evaluating 147 patients with metastatic melanoma treated with immunotherapy and referred for therapy response assessment with ¹⁸F-FDG PET/CT at our institution from January 2010 to August 2022. In total, 201 PET/CT scans performed at various time points were analyzed. IRAEs detected on PET/CT were compared against clinical reference standards, including physical examinations, laboratory tests, and biopsies. Diagnostic performance metrics (sensitivity, specificity, positive predictive value, negative predictive value), and diagnostic yields were calculated. Results: There were 36/147 patients (24.5%) with IRAEs recorded according to standard of reference, with 39 IRAEs in the entire cohort. At time point level, PET/CT identified 36/36 (100%) patients with IRAEs confirmed by the reference standard, while clinical suspicion identified 26/36 (72%) cases. At IRAE level, PET/CT identified 36/39 (92%) of IRAEs confirmed by the reference standard. Thirteen out of 39 (33.3%) cases identified on PET/CT were not suspected clinically but confirmed by the reference standard. The most frequent IRAEs, both suspected clinically and on PET/CT, corresponded to thyroiditis and colitis. Among the PET/CT positive cases, the majority corresponded to grade 2 severity. Conclusion: ¹⁸F-FDG PET/CT is highly effective in detecting IRAEs in patients with metastatic melanoma on immunotherapy, uncovering clinically unsuspected events in up to 33% of cases. These results highlight its important role in early detection, guiding timely interventions, and improving overall outcomes of immunotherapy-related toxicities.

There are few recommendations in Canada to assist clinicians in selecting appropriate biopsy techniques (fine-needle aspiration, core-needle biopsy, vacuum-assisted biopsy, vacuum-assisted excision) and imaging technologies (mammography, ultrasound, magnetic resonance imaging, contrast-enhanced mammography) for biopsy guidance. Limited existing recommendations from other countries do not consider the unique aspects of the Canadian healthcare system. To address this gap, 17 experts participated in a modified Delphi panel to reach consensus on biopsy-related topics and provide recommendations. The panel was comprised of 12 radiologists, 2 pathologists, and 3 surgeons from 6 provinces across Canada. Panelists engaged in two rounds of anonymized voting, with an in-person discussion held between the rounds. The modified Delphi panel adhered to best practices, including establishing consensus definitions prior to voting, utilizing anonymized voting, and abstaining from communication among panelists before the in-person meeting. A rigorous statistical approach was utilized to analyze the points of agreement and disagreement. Consensus findings covered a wide range of topics, including recommendations for initial biopsy technique based on lesion type and imaging modality, patient management or rebiopsy considerations after the initial biopsy, procedural recommendations (i.e., gauge size, number of samples), patient considerations (i.e., drug allergies, pregnancy). Overall, 347 individual items were included in the final analysis, 286 (82%) of which achieved consensus. These consensus recommendations intend to offer general recommendations to help standardize and improve practices across Canada and were endorsed by the Canadian Society of Breast Imaging. However, they should be evaluated in the context of each individual case and emerging evidence.

Environmental degradation and climate change pose an increasingly serious threat to global health, necessitating urgent action to implement environmentally sustainable healthcare practices. Interventional radiology (IR) is a resource-intensive specialty that has not historically emphasized environmental sustainability. This review aims to examine the environmental impact of IR and highlight opportunities for transitioning to more sustainable practices within the IR suite. The environmental impact of IR is assessed in 3 critical domains: (1) energy consumption, (2) waste production, and (3) water pollution. For each domain, actionable strategies are proposed to mitigate environmental harm. Key actions include powering down equipment when not in use, utilizing energysaving modes, minimizing the reliance on single-use items where possible, collaborating with industry to reduce excessive packaging, and implementing recycling programs for waste and iodinated contrast media, along with incorporating environmental sustainability as a quality metric in the departments quality improvement program. Barriers to adopting environmentally sustainable changes include a lack of awareness, financial considerations, and the absence of government, institutional, and industry regulations. Leadership from professional societies and collaboration with industry partners will be essential for driving systemic change. However, individual departments can take action to foster a culture of environmental responsibility and implement sustainable practices.

Background: Adherence of diagnostic accuracy imaging research to the STARD 2015 reporting guideline was assessed at baseline in 2016; on average, only 55% of 30 items were reported. Several knowledge translation strategies have since been implemented by the STARD group. Purpose: The purpose of this study was to evaluate the adherence of diagnostic accuracy studies recently published in imaging journals to STARD 2015, to assess for changes in the level of adherence relative to the baseline study. Methods: We performed an electronic search on MEDLINE for diagnostic accuracy studies, published between May and June of 2024, from a select group of imaging journals. The timespan was modulated to achieve a sample size of 100 to 150 included studies. Overall and item-specific adherence to STARD 2015 was evaluated, in addition to associations with journal of publication, imaging modality, study design, country of corresponding author, imaging subspecialty area, journal impact factor, and journal STARD adoption. Statistical comparison to the baseline study from 2016 was also performed. Poisson Regression and two-tailed student's tests were used to compare STARD adherence relative to variables included in subgroup analysis. Results: In the 126 included studies, average adherence to STARD 2015 was 61% (18.3/30 items; SD = 3.1), improved compared to the baseline study (55%; 16.6/30 items; SD = 2.2; P < .0001). Studies published in higher impact factor journals reported more items than those in lower impact factor journals (20.6 vs 18.4 items, P-value <.0001). There was no significant association between reporting completeness and journal of publication (P = .7), imaging modality (P = .21), country of corresponding author (P = .46), imaging subspecialty (P = .31), and journal STARD adoption status (P = .55). Conclusion: Recently published diagnostic accuracy studies reported more STARD 2015 items than studies published in 2016, but completeness of reporting is still not optimal.

Purpose: Breast arterial calcifications (BAC) are not routinely reported on mammography but are linked to coronary artery calcification (CAC) and cardiovascular disease (CVD) events. We sought to assess primary care provider (PCP) follow-up after BAC and CAC notification and the association between BAC on mammography and CAC on CT. Methods: Participants without known CVD undergoing mammography at a single centre were prospectively recruited over 18 months. BAC were qualitatively scored (none/mild/moderate/severe) by 2 breast radiologists. All participants had research cardiac CT for CAC within 6 months, scored using the Agatston method. Questionnaires collected baseline demographics, risk factors, and follow-up data. Results: 286 participants were included (median age 62 ± 10). Prevalence of BAC was 13% (38/286), 248 had none, 18 mild, 16 moderate, and 4 severe. For CAC: 180 had none, 70 had mild (CAC 1-99), 28 had moderate (CAC 100-399), and 8 had severe (CAC >400). For detecting CAC, BAC presence had 92% specificity (166/180), 23% sensitivity (24/106), and 67% negative predictive value (166/248). Most participants with BAC and CAC (71%, 17/24) were not on lipid-lowering therapy and 63% (15/24) did not believe they had elevated CVD risk. At follow-up (median 202 days), 46% (11/24) with BAC and CAC implemented lifestyle modifications, 92% (22/24) scheduled PCP follow-up, and 56% (10/18) underwent further CV risk assessment following their appointment. One participant with BAC and CAC had a stroke during follow-up. Conclusion: In a prospective cohort without known CVD undergoing mammography, notification of BAC and CAC status prompted high follow-up rates with PCPs and lifestyle modifications.

Ultrasound (US) is an invaluable tool for evaluating the neonatal spine, offering a non-invasive, cost-effective, and radiation-free imaging alternative. This article reviews the indications, techniques, and sonographic features of congenital and acquired spinal pathologies in neonates. Common indications include the evaluation of atypical sacral dimples, spinal anomalies such as tethered cord and syringomyelia, congenital tumours like sacrococcygeal teratomas, and post-traumatic conditions such as spinal hematomas. Detailed sonographic spine anatomy, variants, and ultrasound guidance for interventions like lumbar puncture are also discussed. By enabling early diagnosis, US is crucial in guiding clinical management, especially in pediatric populations with spinal disorders.