Search

Purpose: Large language models (LLMs) like Generative Pre-trained Transformer 4 (GPT-4) can assist in detecting errors in radiology reports, but privacy concerns limit their clinical applicability. This study compares closed-source and privacy-compliant open-source LLMs for detecting common errors in radiology reports., Materials and Methods: A total of 120 radiology reports were compiled (30 each from X-ray, ultrasound, CT, and MRI). Subsequently, 397 errors from five categories (typographical, numerical, findings-impression discrepancies, omission/insertion, interpretation) were inserted into 100 of these reports; 20 reports were left unchanged. Two open-source models (Llama 3-70b, Mixtral 8x22b) and two commercial closed-source (GPT-4, GPT-4o) were tasked with error detection using identical prompts. The Kruskall-Wallis test and paired t-test were used for statistical analysis., Results: Open-source LLMs required less processing time per radiology report than closed-source LLMs (6 ± 2 s vs. 13 ± 4 s; p < 0.001). Closed-source LLMs achieved higher error detection rates than open-source LLMs (GPT-4o: 88% [348/397; 95% CI: 86, 92], GPT-4: 83% [328/397; 95% CI: 80, 87], Llama 3-70b: 79% [311/397; 95% CI: 76, 83], Mixtral 8x22b: 73% [288/397; 95% CI: 68, 77]; p < 0.001). Numerical errors (88% [67/76; 95% CI: 82, 93]) were detected significantly more often than typographical errors (75% [65/86; 95% CI: 68, 82]; p = 0.02), discrepancies between findings and impression (73% [73/101; 95% CI: 67, 80]; p < 0.01), and interpretation errors (70% [50/71; 95% CI: 62, 78]; p = 0.001)., Conclusion: Open-source LLMs demonstrated effective error detection, albeit with comparatively lower accuracy than commercial closed-source models, and have potential for clinical applications when deployed via privacy-compliant local hosting solutions., Key Points: Question Can privacy-compliant open-source large language models (LLMs) match the error-detection performance of commercial non-privacy-compliant closed-source models in radiology reports? Findings Closed-source LLMs achieved slightly higher accuracy in detecting radiology report errors than open-source models, with Llama 3-70b yielding the best results among the open-source models. Clinical relevance Open-source LLMs offer a privacy-compliant alternative for automated error detection in radiology reports, improving clinical workflow efficiency while ensuring patient data confidentiality. Further refinement could enhance their accuracy, contributing to better diagnosis and patient care., Competing Interests: Compliance with ethical standards. Guarantor: The scientific guarantor of this publication is Alexander Isaak, MD. Conflict of interest: The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article. Statistics and biometry: No complex statistical methods were necessary for this paper. Informed consent: Informed consent was not required as no patient-identifying information was used. Ethical approval: After consultation, specific approval from the institutional review board was not required as no patient-identifying information was used. Study subjects or cohorts overlap: Not applicable. Methodology: Experimental, (© 2025. The Author(s).)

Objectives: To investigate whether proton density fat fraction (PDFF) measurements using a six-echo modified Dixon sequence can help to differentiate between benign and malignant vertebral bone marrow lesions., Methods: Sixty-six patients were prospectively enrolled in our study. In addition to conventional MRI at 3.0-Tesla including at least sagittal T2-weighted/spectral attenuated inversion recovery and T1-weighted sequences, all patients underwent a sagittal six-echo modified Dixon sequence of the spine. The mean PDFF was calculated using regions of interest and compared between vertebral lesions. A cut-off value of 6.40% in PDFF was determined by receiver operating characteristic curves and used to differentiate between malignant (< 6.40%) and benign (≥ 6.40%) vertebral lesions., Results: There were 77 benign and 44 malignant lesions. The PDFF of malignant lesions was statistically significant lower in comparison with benign lesions (p < 0.001) and normal vertebral bone marrow (p < 0.001). The areas under the curves (AUC) were 0.97 for differentiating benign from malignant lesions (p < 0.001) and 0.95 for differentiating acute vertebral fractures from malignant lesions (p < 0.001). This yielded a diagnostic accuracy of 96% in the differentiation of both benign lesions and acute vertebral fractures from malignancy., Conclusion: PDFF derived from six-echo modified Dixon allows for differentiation between benign and malignant vertebral lesions with a high diagnostic accuracy., Key Points: • Establishing a diagnosis of indeterminate vertebral lesions is a common clinical problem • Benign bone marrow processes may mimic the signal alterations observed in malignancy • PDFF differentiates between benign and malignant lesions with a high diagnostic accuracy • PDFF of non-neoplastic vertebral lesions is significantly higher than that of malignancy • PDFF from six-echo modified Dixon may help avoid potentially harmful bone biopsy.

Objectives: To investigate the diagnostic value of cardiac magnetic resonance (CMR) feature-tracking (FT) myocardial strain analysis in patients with suspected acute myocarditis and its association with myocardial oedema., Methods: Forty-eight patients with suspected acute myocarditis and 35 control subjects underwent CMR. FT CMR analysis of systolic longitudinal (LS), circumferential (CS) and radial strain (RS) was performed. Additionally, the protocol allowed for the assessment of T1 and T2 relaxation times., Results: When compared with healthy controls, myocarditis patients demonstrated reduced LS, CS and RS values (LS: -19.5 ± 4.4% vs. -23.6 ± 3.1%, CS: -23.0 ± 5.8% vs. -27.4 ± 3.4%, RS: 28.9 ± 8.5% vs. 32.4 ± 7.4%; P < 0.05, respectively). LS (T1: r = 0.462, P < 0.001; T2: r = 0.436, P < 0.001) and CS (T1: r = 0.429, P < 0.001; T2: r = 0.467, P < 0.001) showed the strongest correlations with T1 and T2 relaxations times. Area under the curve of LS (0.79) was higher compared with those of CS (0.75; P = 0.478) and RS (0.62; P = 0.008)., Conclusions: FT CMR myocardial strain analysis might serve as a new tool for assessment of myocardial dysfunction in the diagnostic work-up of patients suspected of having acute myocarditis. Especially, LS and CS show a sufficient diagnostic performance and were most closely correlated with CMR parameters of myocardial oedema., Key Points: • Myocardial strain measures are considerably reduced in patients with suspected myocarditis. • Myocardial strain measures can sufficiently discriminate between diseased and healthy patients. • Myocardial strain measures show basic associations with the extent of myocardial oedema/inflammation.