Performance of an Open-Source Large Language Model in Extracting Information from Free-Text Radiology Reports.

Purpose To assess the performance of a local open-source large language model (LLM) in various information extraction tasks from real-life emergency brain MRI reports. Materials and Methods All consecutive emergency brain MRI reports written in 2022 from a French quaternary center were retrospectively reviewed. Two radiologists identified MRI scans that were performed in the emergency department for headaches. Four radiologists scored the reports' conclusions as either normal or abnormal. Abnormalities were labeled as either headache-causing or incidental. Vicuna (LMSYS Org), an open-source LLM, performed the same tasks. Vicuna's performance metrics were evaluated using the radiologists' consensus as the reference standard. Results Among the 2398 reports during the study period, radiologists identified 595 that included headaches in the indication (median age of patients, 35 years [IQR, 26-51 years]; 68% [403 of 595] women). A positive finding was reported in 227 of 595 (38%) cases, 136 of which could explain the headache. The LLM had a sensitivity of 98.0% (95% CI: 96.5, 99.0) and specificity of 99.3% (95% CI: 98.8, 99.7) for detecting the presence of headache in the clinical context, a sensitivity of 99.4% (95% CI: 98.3, 99.9) and specificity of 98.6% (95% CI: 92.2, 100.0) for the use of contrast medium injection, a sensitivity of 96.0% (95% CI: 92.5, 98.2) and specificity of 98.9% (95% CI: 97.2, 99.7) for study categorization as either normal or abnormal, and a sensitivity of 88.2% (95% CI: 81.6, 93.1) and specificity of 73% (95% CI: 62, 81) for causal inference between MRI findings and headache. Conclusion An open-source LLM was able to extract information from free-text radiology reports with excellent accuracy without requiring further training. Keywords: Large Language Model (LLM), Generative Pretrained Transformers (GPT), Open Source, Information Extraction, Report, Brain, MRI Supplemental material is available for this article. Published under a CC BY 4.0 license. See also the commentary by Akinci D'Antonoli and Bluethgen in this issue.

Depolarization-induced bursts of miniature synaptic currents in individual synapses of developing cerebellum.

In central synapses, spontaneous transmitter release observed in the absence of action potential firing is often considered as a random process lacking time or space specificity. However, when studying miniature glutamatergic currents at cerebellar synapses between parallel fibers and molecular layer interneurons, we found that these currents were sometimes organized in bursts of events occurring at high frequency (about 30 Hz). Bursts displayed homogeneous quantal size amplitudes. Furthermore, in the presence of the desensitization inhibitor cyclothiazide, successive events within a burst displayed quantal amplitude occlusion. Based on these findings, we conclude that bursts originate in individual synapses. Bursts were enhanced by increasing either the external potassium concentration or the external calcium concentration, and they were strongly inhibited when blocking voltage-gated calcium channels by cadmium. Bursts were prevalent in elevated potassium concentration during the formation of the molecular layer but were infrequent later in development. Since postsynaptic AMPA receptors are largely calcium permeant in developing parallel fiber-interneuron synapses, we propose that bursts involve presynaptic calcium transients implicating presynaptic voltage-gated calcium channels, together with postsynaptic calcium transients implicating postsynaptic AMPA receptors. These simultaneous pre- and postsynaptic calcium transients may contribute to the formation and/or stabilization of synaptic connections.

Flux coupling, not specificity, shapes the transport and phylogeny of SLC6 glycine transporters.

ATB[Formula: see text] (SLC6A14) is a member of the amino acid transporter branch of the SLC6 family along with GlyT1 (SLC6A9) and GlyT2 (SLC6A5), two glycine-specific transporters coupled to 2:1 and 3:1 Na[Formula: see text]:Cl[Formula: see text], respectively. In contrast, ATB[Formula: see text] exhibits broad substrate specificity for all neutral and cationic amino acids, and its ionic coupling remains unsettled. Using the reversal potential slope method, we demonstrate a 3:1:1 Na[Formula: see text]:Cl[Formula: see text]:Gly stoichiometry for ATB[Formula: see text] that is consistent with its 2.1 e/Gly charge coupling. Like GlyT2, ATB[Formula: see text] behaves as a unidirectional transporter with virtually no glycine efflux at negative potentials after uptake, except by heteroexchange as remarkably shown by leucine activation of NMDARs in Xenopus oocytes coexpressing both membrane proteins. Analysis and computational modeling of the charge movement of ATB[Formula: see text] reveal a higher affinity for sodium in the absence of substrate than GlyT2 and a gating mechanism that locks Na[Formula: see text] into the apo-transporter at depolarized potentials. A 3:1 Na[Formula: see text]:Cl[Formula: see text] stoichiometry justifies the concentrative transport properties of ATB[Formula: see text] and explains its trophic role in tumor growth, while rationalizing its phylogenetic proximity to GlyT2 despite their extreme divergence in specificity.

Test-retest reproducibility of cardiac magnetic resonance imaging in healthy mice at 7-Tesla: effect of anesthetic procedures.

Cardiac magnetic resonance (CMR) has emerged as a powerful tool for in vivo assessments of cardiac parameters in experimental animal models of cardiovascular diseases, but its reproducibility in this setting remains poorly explored. To address this issue, we investigated the test-retest reproducibility of preclinical cardiac magnetic resonance imaging (CMR) at 7 Tesla in healthy C57BL/6 mice, including an analysis of the impact of different anesthetic procedures (isoflurane or pentobarbital). We also analyzed the intra-study reproducibility and the intra- and inter-observer post-processing reproducibility of CMR images. Test-retest reproducibility was high for left ventricular parameters, especially with the isoflurane anesthetic procedure, whereas right ventricular parameters and deformation measurements were less reproducible, mainly due to physiological variability. Post-processing reproducibility of CMR images was high both within and between observers. These results highlight that anesthetic procedures might influence CMR test-retest reproducibility, an important ethical consideration for longitudinal studies in rodent models of cardiomyopathy to limit the number of animals used.