Search

Purpose: The MR-guided radiotherapy (MRgRT) images on the integrated MRI and linear accelerator (MR-LINAC) might facilitate radiomics analysis for longitudinal treatment response assessment. However, the reliability of MRgRT radiomics features is largely unknown. This study aims to investigate MRgRT radiomics feature reliability acquired using a standardized 3D-T2W-TSE sequence in terms of repeatability, reproducibility, and within-subject feature agreement on a 1.5T MR-simulator and a 1.5T MR-LINAC for prostate cancer (PC)., Methods: Twenty-six consecutive PC patients who underwent one MRI-simulator scan and two MR-LINAC scans before dose delivery were retrospectively included. The three MRI datasets were rigidly co-registered. 1023 first-order and texture radiomics features were extracted with different intensity bin widths for each scan in the manually segmented clinical target volume (CTV) and planning target volume (PTV) by an experienced radiation oncologist. Intraclass correlation coefficient (ICC) was used to evaluate feature repeatability between MR-LINAC scans and reproducibility between MRI-simulator and MR-LINAC scans. The within-subject feature value agreements were evaluated using Bland-Altman analysis. The impact of inter-observer segmentation on the radiomics feature reliability was also examined based on the second manual segmentation of CTV and PTV by an MRI researcher., Results: Based on the segmentation by the radiation oncologist and the default bin width of 25, 9.6%, 24.1%, 49.6%, and 16.8% of the total 1023 features exhibited excellent (ICC > 0.9), good (0.9 > ICC > 0.75), moderate (0.75 > ICC > 0.5), and poor (ICC < 0.5) repeatability in the CTV, and 9.2%, 26.8%, 50.5%, and 13.5% in the PTV, respectively. For reproducibility, the corresponding feature percentages were 8.9%, 19.7%, 41.9%, and 29.6% in the CTV, and 8.4%, 17.8%, 47.9%, and 26% in the PTV. Feature reliability was not notably influenced by intensity bin width for discretization. BA analysis revealed wide 95% limit-of-agreements and substantial biases of feature values between CTV and PTV and between any two MRI scans. The features even with excellent ICC were still subjected to considerable inter-scan feature variations in each individual subject. The analysis on the second segmentation by the MRI researcher showed insignificantly different feature repeatability and reproducibility in terms of ICC values., Conclusions: Only a small proportion of features exhibited excellent/good repeatability and reproducibility, highlighting the importance of reliable MRgRT feature selection. The within-subject feature values were subjected to considerable inter-scan variations, imposing a challenge on the determination of the smallest detectable change in future MRgRT delta-radiomics studies., (© 2021 American Association of Physicists in Medicine.)

High dimensional neuroimaging datasets and machine learning have been used to estimate and predict domain-specific cognition, but comparisons with simpler models composed of easy-to-measure variables are limited. Regularization methods in particular may help identify regions-of-interest related to domain-specific cognition. Using data from the Northern Manhattan Study, a cohort study of mostly Hispanic older adults, we compared three models estimating domain-specific cognitive performance: sociodemographics and APOE ε4 allele status (basic model), the basic model and MRI markers, and a model with only MRI markers. We used several machine learning methods to fit our regression models: elastic net, support vector regression, random forest, and principal components regression. Model performance was assessed with the RMSE, MAE, and R 2 statistics using 5-fold cross-validation. To assess whether prediction models with imaging biomarkers were more predictive than prediction models built with randomly generated biomarkers, we refit the elastic net models using 1000 datasets with random biomarkers and compared the distribution of the RMSE and R 2 in models using these random biomarkers to the RMSE and R 2 from observed models. Basic models explained ~ 31-38% of the variance in domain-specific cognition. Addition of MRI markers did not improve estimation. However, elastic net models with only MRI markers performed significantly better than random MRI markers (one-sided P < .05) and yielded regions-of-interest consistent with previous literature and others not previously explored. Therefore, structural brain MRI markers may be more useful for etiological than predictive modeling., (© 2020. Springer Science+Business Media, LLC, part of Springer Nature.)

Background and Purpose: Cognition and education influence functional trajectories, but whether associations differ with subclinical brain infarcts (SBI) or white matter hyperintensity volume (WMHV) is unknown. We hypothesized that SBI and WMHV moderated relationships between cognitive performance and education and functional trajectories., Methods: A total of 1290 stroke-free individuals underwent brain magnetic resonance imaging and were followed for 7.3 years (mean) with annual functional assessments with the Barthel index (range, 0-100). Magnetic resonance imaging measurements included pathology-informed SBI (PI-SBI) and WMHV (% total cranial volume). Generalized estimating equation models tested associations between magnetic resonance imaging variables and baseline Barthel index and change in Barthel index, adjusting for demographic, vascular, cognitive, and social risk factors, and stroke and myocardial infarction during follow-up. We tested interactions among education level, baseline cognitive performance (Mini-Mental State score), and functional trajectories and ran models stratified by levels of magnetic resonance imaging variables., Results: Mean age was 70.6 (SD, 9.0) years; 19% had PI-SBI, and mean WMHV was 0.68%. Education did not modify associations between cognition and functional trajectories. PI-SBI modified associations between cognition and functional trajectories ( P =0.04) with a significant protective effect of better cognition on functional decline seen only in those without PI-SBI. There was no significant interaction for WMHV ( P =0.8). PI-SBI, and greater WMHV, were associated with 2- to 3-fold steeper functional decline, holding cognition constant., Conclusions: PI-SBI moderated the association between cognition and functional trajectories, with 3-fold greater decline among those with PI-SBI (compared with no PI-SBI) and normal baseline cognition. This highlights the strong and independent association between subclinical markers and patient-centered trajectories over time., (© 2018 American Heart Association, Inc.)

Multiple sclerosis (MS) patients occasionally present with homeostatic disturbances suggestive of hypothalamic dysfunction; however, the hypothalamus often appears normal on imaging, apart from confirmation in a few necropsy studies. In this article we describe a Chinese woman with atypical clinical presentation. She presented with relapsing symptoms and signs consistent with hypothalamic dysfunction including hyperprolactinema, syndrome of inappropriate secretion of anti-diuretic hormone (SIADH), hypersomnolence and temperature dysregulation. Serial MRI depicted the atypical reversible changes in the hypothalamus, correlating with the observed homeostatic abnormalities.

The aim of the study was to investigate the feasibility of using digital subtraction in contrast-enhanced MR imaging of the brain to reduce the MR contrast dosage without jeopardizing patient care. Fifty-two patients with intracranial lesions, either intra-axial or extra-axial, detected by computerized tomography were selected for contrast-enhanced MR imaging with half-dose and full-dose of gadopentetate dimeglumine. The half-dose unsubtracted, full-dose unsubtracted, and half-dose subtracted MR images were visually assessed by counting the number of enhancing brain lesions in the images and quantitatively analyzed by computing their lesion contrast-to-background ratios (CBR). The visual conspicuity of the half-dose subtracted MR images was comparable to that of the full-dose unsubtracted MR images ( p>0.05), whereas the CBR of the half-dose subtracted images was approximately two to three times higher than that of the full-dose unsubtracted images. The half-dose subtracted T1-weighted spin-echo images might be able to replace the conventional standard-dose T1-weighted spin-echo images in MR imaging of the brain.

Late temporal lobe necrosis is a well-known and serious complication in patients with nasopharyngeal carcinoma (NPC) following radiotherapy. Owing to the close proximity to the skull base, the medial temporal lobes are inevitably included in the target volume of irradiation. Patients with NPC provide a unique opportunity in study of delay radiation effect in normal human brain. The objective of this study was to evaluate late temporal lobe radiation injury by combined multi-section diffusion weighted and perfusion weighted MR imaging. We prospectively studied 16 patients with typical clinical symptoms of late temporal lobe necrosis or other abnormalities in the temporal lobes incidentally detected by conventional MR imaging. All patients had a previous history of radiotherapy for histologically proven NPC. Conventional T1- and T2-weighted images, fast gradient echo with echo-planar diffusion-weighted and perfusion-weighted MR imaging were performed. Apparent diffusion coefficient (ADC) map and relative cerebral blood volume (rCBV) map were computed via commercially available software. MR diffusion and perfusion images were then analyzed and graded by two independent observers with focusing on the diffusion and perfusion mismatch. The temporal lobe lesions displayed marked high diffusion on the ADC map. The rCBV map also revealed marked hypoperfusion in these temporal lobe lesions in all patients. The areas of abnormality on the rCBV map were significantly larger than the lesions on the ADC map in 14 patients (observer 1) and 13 patients (observer 2). Since late temporal lobe necrosis is probably caused by damage of the endothelium of vessels and ischemia, perfusion and diffusion mismatch might imply injured tissue but potentially salvageable brain tissue. A mismatch may be potentially used to predict the response to treatment in-patients with late temporal lobe necrosis.

Spontaneous intracranial hypotension is a rare phenomenon characterized by postural headache, neck rigidity, nausea and vomiting. Imaging findings on magnetic resonance imaging (MRI) is characteristic with diffuse intracranial pachymeningeal thickening and enhancement following intravenous gadolinium. We present a case of spontaneous intracranial hypotension with two unusual imaging findings; pachymeningeal enhancement of the spinal canal and enlargement of the pituitary gland in addition to the diffuse intracranial pachymeningeal enhancement. In this case report, we will discuss the clinical features, MRI findings and underlying pathophysiology of this rare condition.

Background: We evaluated the ability of diffusion-weighted (DW) magnetic resonance (MR) imaging to differentiate between hepatic abscess and cystic or necrotic liver tumor., Methods: DW MR imaging was performed in 18 patients who had liver masses with large cystic or necrotic cavities detected by computed tomography or ultrasonography. The final diagnoses were pyogenic abscess (five cases), fungal abscess (one), hepatocellular carcinoma (five), cystic metastasis (four), and hepatic cyst (three). Signal intensities on the DW images and apparent diffusion coefficient (ADC) maps were noted. ADCs in all lesions were calculated and compared with Student's t test., Results: All lesions, except hepatic cysts, showed either heterogeneous or homogeneous rim enhancement on postgadolinium images. All abscess cavities showed hyperintensity on DW images and hypointensity on ADC maps. Conversely, the cystic or necrotic portions of all tumors showed hypointensity on DW images and hyperintensity on ADC maps. The mean ADCs (mm(2)/s) were 0.67 +/- 0.35 x 10(-3) in hepatic abscess, 2.65 +/- 0.49 x 10(-3) in cystic or necrotic tumor, 2.93 +/- 0.52 x 10(-3) in hepatic cyst, and 1.98 +/- 0.37 x 10(-3) in normal liver parenchyma., Conclusion: DW MR imaging, with the help of ADCs, may be a useful noninvasive imaging technique in differentiating hepatic abscess from cystic or necrotic tumor.

We examined nine patients with histologically proven nasopharyngeal carcinoma (NPC), treated with radiotherapy, using dynamic susceptibility contrast MRI (DSC-MRI). In eight there was clinical evidence of radionecrosis of the temporal lobe, and one was examined for local recurrence in the nasopharynx. All patients had either high-signal finger-like or cystic lesions in the temporal lobes on T2-weighted images. Heterogeneous contrast enhancement occurred in all patients. Relative regional cerebral blood volume (rrCBV) mapping revealed marked hypoperfusion in all patients. One underwent bilateral temporal lobectomy and radionecrosis was confirmed histologically.

Spinal cord ganglioglioma is a rare tumor most often encountered in the first three decades. Scanty computed tomography (CT) reports on the tumor describe it as a hypodense or cerebrospinal fluid (CSF) dense area with little contrast enhancement despite its solid nature. We report two cases of spinal ganglioglioma both involving almost the whole spinal cord. On magnetic resonance imaging (MRI), the tumors appear hypointense to the spinal cord on T1 and hyperintense to the cord on T2 images, and were mainly solid at exploration. It is important to recognize these tumors as long survival can be achieved after surgical resection.