GPT-4 (ChatGPT; OpenAI) matched the performance of radiologists in detecting errors in radiology reports and has the potential to reduce processing time and cost.

Background: To investigate the feasibility of the large language model (LLM) ChatGPT for classifying liver lesions according to the Liver Imaging Reporting and Data System (LI-RADS) based on MRI reports, and to compare classification performance on structured vs. unstructured reports.Methods: LI-RADS classifiable liver lesions were included from German written structured and unstructured MRI reports with report of size, location, and arterial phase contrast enhancement as minimum inclusion requirements. The findings sections of the reports were propagated to ChatGPT (GPT-3.5), which was instructed to determine LI-RADS scores for each classifiable liver lesion. Ground truth was established by two radiologists in consensus. Agreement between ground truth and ChatGPT was assessed with Cohen's kappa. Test-retest reliability was assessed by passing a subset of n = 50 lesions five times to ChatGPT, using the intraclass correlation coefficient (ICC).Results: 205 MRIs from 150 patients were included. The accuracy of ChatGPT at determining LI-RADS categories was poor (53% and 44% on unstructured and structured reports). The agreement to the ground truth was higher (k = 0.51 and k = 0.44), the mean absolute error in LI-RADS scores was lower (0.5 ± 0.5 vs. 0.6 ± 0.7, p < 0.05), and the test-retest reliability was higher (ICC = 0.81 vs. 0.50), in free-text compared to structured reports, respectively, although structured reports comprised the minimum required imaging features significantly more frequently (Chi-square test, p < 0.05).Conclusions: ChatGPT attained only low accuracy when asked to determine LI-RADS scores from liver imaging reports. The superior accuracy and consistency throughout free-text reports might relate to ChatGPT's training process.Clinical relevance statement: Our study indicates both the necessity of optimization of LLMs for structured clinical data input and the potential of LLMs for creating machine-readable labels based on large free-text radiological databases.

To investigate the potential of combining Compressed Sensing (CS) and a newly developed AI-based super resolution reconstruction prototype consisting of a series of convolutional neural networks (CNN) for a complete five-minute 2D knee MRI protocol. In this prospective study, 20 volunteers were examined using a 3T-MRI-scanner (Ingenia Elition X, Philips). Similar to clinical practice, the protocol consists of a fat-saturated 2D-proton-density-sequence in coronal, sagittal and transversal orientation as well as a sagittal T1-weighted sequence. The sequences were acquired with two different resolutions (standard and low resolution) and the raw data reconstructed with two different reconstruction algorithms: a conventional Compressed SENSE (CS) and a new CNN-based algorithm for denoising and subsequently to interpolate and therewith increase the sharpness of the image (CS-SuperRes). Subjective image quality was evaluated by two blinded radiologists reviewing 8 criteria on a 5-point Likert scale and signal-to-noise ratio calculated as an objective parameter. The protocol reconstructed with CS-SuperRes received higher ratings than the time-equivalent CS reconstructions, statistically significant especially for low resolution acquisitions (e.g., overall image impression: 4.3 ± 0.4 vs. 3.4 ± 0.4, p < 0.05). CS-SuperRes reconstructions for the low resolution acquisition were comparable to traditional CS reconstructions with standard resolution for all parameters, achieving a scan time reduction from 11:01 min to 4:46 min (57 %) for the complete protocol (e.g. overall image impression: 4.3 ± 0.4 vs. 4.0 ± 0.5, p < 0.05). The newly-developed AI-based reconstruction algorithm CS-SuperRes allows to reduce scan time by 57% while maintaining unchanged image quality compared to the conventional CS reconstruction.

Background: The latest advancement of artificial intelligence (AI) is generative pretrained transformer large language models (LLMs). They have been trained on massive amounts of text, enabling humanlike and semantical responses to text-based inputs and requests. Foreshadowing numerous possible applications in various fields, the potential of such tools for medical data integration and clinical decision-making is not yet clear. Objective: In this study, we investigate the potential of LLMs in report-based medical decision-making on the example of acute ischemic stroke (AIS), where clinical and image-based information may indicate an immediate need for mechanical thrombectomy (MT). The purpose was to elucidate the feasibility of integrating radiology report data and other clinical information in the context of therapy decision-making using LLMs. Methods: A hundred patients with AIS were retrospectively included, for which 50% (50/100) was indicated for MT, whereas the other 50% (50/100) was not. The LLM was provided with the computed tomography report, information on neurological symptoms and onset, and patients’ age. The performance of the AI decision-making model was compared with an expert consensus regarding the binary determination of MT indication, for which sensitivity, specificity, and accuracy were calculated. Results: The AI model had an overall accuracy of 88%, with a specificity of 96% and a sensitivity of 80%. The area under the curve for the report-based MT decision was 0.92. Conclusions: The LLM achieved promising accuracy in determining the eligibility of patients with AIS for MT based on radiology reports and clinical information. Our results underscore the potential of LLMs for radiological and medical data integration. This investigation should serve as a stimulus for further clinical applications of LLMs, in which this AI should be used as an augmented supporting system for human decision-making.

Objectives: The purpose of this study was to assess whether a 3-min 2D knee protocol can meet the needs for clinical application if using a SuperResolution reconstruction approach. Methods: In this prospective study, a total of 20 volunteers underwent imaging of the knee using a 3T MRI scanner (Philips Ingenia Elition X 3.0T, Philips). The imaging protocol, consisting of a fat-saturated 2D proton density sequence in coronal, sagittal, and transverse orientations, as well as a sagittal T1-weighted sequence, was acquired with standard and ultra-low resolution. The standard sequences were reconstructed using an AI-assisted Compressed SENSE method (SmartSpeed). The ultra-low-resolution sequences have been reconstructed using a vendor-provided prototype. Four experienced readers (two radiologists and two orthopedic surgeons) evaluated the sequences for image quality, anatomical structures, and incidental pathologies. The consensus evaluation of two different experienced radiologists specialized in musculoskeletal imaging served as the gold standard. Results: The acquisition time for the entire protocol was 11:01 min for standard resolution and 03:36 min for ultra-low resolution. In the overall assessment, CS-SuperRes-reconstructed sequences showed slightly improved accuracy and increased specificity compared to the standard CS-AI method (0.87 vs. 0.86 and 0.9 vs. 0.87, respectively), while the standard method exhibited a higher sensitivity (0.73 vs. 0.57). Overall, 24 out of 40 pathologies were detected in the ultra-low-resolution images compared to 26 in the standard images. Conclusions: The CS-SuperRes method enables a 2D knee protocol to be completed in 3 min, with improved accuracy compared to the clinical standard.