Search

Artificial intelligence (AI) models often face performance drops after deployment to external datasets. This study evaluated the potential of a novel data augmentation framework based on generative adversarial networks (GANs) that creates synthetic patient image data for model training to improve model generalizability. Model development and external testing were performed for a given classification task, namely the detection of new fluid-attenuated inversion recovery lesions at MRI during longitudinal follow-up of patients with multiple sclerosis (MS). An internal dataset of 669 patients with MS ( n = 3083 examinations) was used to develop an attention-based network, trained both with and without the inclusion of the GAN-based synthetic data augmentation framework. External testing was performed on 134 patients with MS from a different institution, with MR images acquired using different scanners and protocols than images used during training. Models trained using synthetic data augmentation showed a significant performance improvement when applied on external data (area under the receiver operating characteristic curve [AUC], 83.6% without synthetic data vs 93.3% with synthetic data augmentation; P = .03), achieving comparable results to the internal test set (AUC, 95.0%; P = .53), whereas models without synthetic data augmentation demonstrated a performance drop upon external testing (AUC, 93.8% on internal dataset vs 83.6% on external data; P = .03). Data augmentation with synthetic patient data substantially improved performance of AI models on unseen MRI data and may be extended to other clinical conditions or tasks to mitigate domain shift, limit class imbalance, and enhance the robustness of AI applications in medical imaging. Keywords: Brain, Brain Stem, Multiple Sclerosis, Synthetic Data Augmentation, Generative Adversarial Network Supplemental material is available for this article. © RSNA, 2024.

Objectives: Quantitative T1 relaxometry is the benchmark in imaging potential gadolinium deposition and known to be superior to semiquantitative signal intensity ratio analyses. However, T1 relaxometry studies are rare, commonly limited to a few target structures, and reported results are inconsistent.We systematically investigated quantitative T1 relaxation times (qT1) of a variety of brain nuclei after serial application of gadobutrol., Materials and Methods: Retrospectively, qT1 measurements were performed in a patient cohort with a mean number of 11 gadobutrol applications (n = 46) and compared with a control group with no prior gadolinium-based contrast agent administration (n = 48). The following target structures were evaluated: dentate nucleus, globus pallidus, thalamus, hippocampus, putamen, caudate, amygdala, and different white matter areas. Subsequently, multivariate regression analysis with adjustment for age, presence of brain metastases and previous cerebral radiotherapy was performed., Results: No assessed site revealed a significant correlation between qT1 and number of gadobutrol administrations in multivariate regression analysis. However, a significant negative correlation between qT1 and age was found for the globus pallidus as well as anterior and lateral thalamus (P < 0.05 each)., Conclusions: No T1 relaxation time shortening due to gadobutrol injection was found in any of the assessed brain structures after serial administration of 11 doses of gadobutrol.

Background: The "glymphatic system" (GS), a brain-wide network of cerebrospinal fluid microcirculation, supplies a pathway through and out of the central nervous system (CNS); malfunction of the system is implicated in a variety of neurological disorders. In this exploratory study, we analyzed the potential of a new imaging approach that we coined delayed T2-weighted gadolinium-enhanced imaging to visualize the GS in vivo., Methods: Heavily T2-weighted fluid-attenuated inversion recovery (hT2w-FLAIR) magnetic resonance imaging was obtained before, and 3 hours and 24 hours after intravenous gadolinium-based contrast agent (GBCA) application in 33 neurologically healthy patients and 7 patients with an impaired blood-brain barrier (BBB) due to cerebral metastases. Signal intensity (SI) was determined in various cerebral fluid spaces, and white matter hyperintensities were quantified by applying the Fazekas scoring system., Findings: Delayed hT2w-FLAIR showed GBCA entry into the CNS via the choroid plexus and the ciliary body, with GBCA drainage along perineural sheaths of cranial nerves and along perivascular spaces of penetrating cortical arteries. In all patients and all sites, a significant SI increase was found for the 3 hours and 24 hours time points compared with baseline. Although no significant difference in SI was found between neurologically healthy patients and patients with an impaired BBB, a significant positive correlation between Fazekas scoring system and SI increase in the perivascular spaces 3 hours post injection was shown., Interpretation: Delayed T2-weighted gadolinium-enhanced imaging can visualize the GBCA pathway into and through the GS. Presence of GBCAs within the GS might be regarded as part of the natural excretion process and should not be mixed up with gadolinium deposition. Rather, the correlation found between deep white matter hyperintensities, an imaging sign of vascular dementia, and GS functioning demonstrated feasibility to exploit the pathway of GBCAs through the GS for diagnostic purposes.

Purpose To determine the effect of at least five serial injections of the macrocyclic gadolinium-based contrast agent (GBCA) gadoterate meglumine on the signal intensity (SI) of the dentate nucleus (DN) of the pediatric brain on nonenhanced T1-weighted magnetic resonance (MR) images. Materials and Methods In this retrospective, institutional review board-approved study, 41 pediatric patients (age range, 3-17 years) who were imaged in at least five consecutive 1.5-T MR examinations with the exclusive use of gadoterate meglumine (plus a final additional nonenhanced MR imaging examination) were evaluated. SI ratio differences between the first and last MR examination were calculated for DN-to-pons and DN-to-middle cerebellar peduncle (MCP) ratios in a region-of-interest-based analysis, and one-sample t tests were used to examine if the SI ratio differences differed from 0. Bayes factors were calculated to quantify the strength of evidence for each test. Results Patients underwent a mean of 8.6 ± 3.9 GBCA administrations (mean accumulated dose, 32.07 mmol ± 17.62, with an average of 16.7 weeks ± 7.9 between every administration). Both ratio differences did not differ significantly from 0 (DN-to-pons ratio: -0.0012 ± 0.0101, P = .436; DN-to-MCP ratio: 0.0007 ± 0.0088, P = .604), and one-sided Bayes factors provided substantial evidence against an SI ratio increase (0.10 for DN-to-pons ratio; 0.27 for DN-to-MCP ratio). Conclusion No increase of the SI in the DN was found after a mean of 8.6 serial injections of the macrocyclic GBCA gadoterate meglumine in pediatric patients, confirming previous studies that did not find this effect after serial injections of macrocyclic GBCAs in adults. © RSNA, 2017.

Purpose To determine the effect of more than 20 serial injections of macrocyclic gadolinium-based contrast agents (GBCAs) on the signal intensity (SI) of the dentate nucleus (DN) on unenhanced T1-weighted magnetic resonance (MR) images. Materials and Methods In this retrospective, institutional review board-approved study, 33 patients who underwent at least 20 consecutive MR imaging examinations (plus an additional MR imaging for reference) with the exclusive use of macrocyclic GBCAs gadoterate meglumine and gadobutrol were analyzed. SI ratio differences were calculated for DN-to-pons and DN-to-middle cerebellar peduncle (MCP) ratios by subtracting the SI ratio at the first MR imaging examination from the SI ratio at the last MR imaging examination. One-sample t tests were used to examine if the SI ratio differences differed from 0, and Bayes factors were calculated to quantify the strength of evidence for each test. Results Patients underwent a mean of 23.03 ± (standard deviation) 4.20 GBCA administrations (mean accumulated dose, 491.21 mL ± 87.04 of a 0.5 M GBCA solution) with an average of 12.09 weeks ± 2.16 between every administration. Both ratio differences did not differ significantly from 0 (DN-to-pons ratio: -0.0032 ± 0.0154, P = .248; DN-to-MCP ratio: -0.0011 ± 0.0093, P = .521), and one-sided Bayes factors provided substantial to strong evidence against an SI ratio increase (Bayes factor for DN-to-pons ratio = 0.09 and that for DN-to-MCP ratio = 0.12). Conclusion The study indicates that 20 or more serial injections of macrocyclic GBCAs administered with on average 3 months between each injection are not associated with an SI increase in the DN. © RSNA, 2016.

We examined the role of total body magnetic resonance imaging (TB-MRI)-governed involved compartment irradiation (ICI) and high-dose chemotherapy (HDC), followed by stem cell rescue (SCR) in patients with high-risk Ewing tumors (ETs) with multiple primary bone metastases (high-risk ET-MBM). Eleven patients with high-risk ET-MBM receiving initial assessment of involved bones by TB-MRI were registered from 1995 to 2000 (group A). In all, 6 patients out of 11 had additional lung disease at initial diagnosis; all had multifocal bone disease with more than three bones involved. After systemic induction with etoposide, vincristine, adriamycin (doxorubicin), ifosfamide, and actinomycin D (EVAIA) or VAIA chemotherapy, ICI of all sites positive by TB-MRI was administered, followed by HDC and SCR. A second group matched for observation period and consisting of 26 patients with more than three involved bones at diagnosis was treated with the European Intergroup Cooperative Ewing Sarcoma Study-92 (EICESS-92) protocol (group B). These patients did not receive TB-MRI and consequently did not receive TB-MRI-governed ICI, or HDC and SCR. Survival in group A vs group B was 45 vs 8% at 5 years and 27 vs 8% at 10 years after diagnosis (log rank and Breslow: P<0.005). We conclude that TB-MRI-governed ICI followed by HDC and SCR in ET-MBM is feasible and warrants further evaluation in prospective studies. [ABSTRACT FROM AUTHOR]