Deep Learning-Driven Transformation: A Novel Approach for Mitigating Batch Effects in Diffusion MRI Beyond Traditional Harmonization.

BACKGROUND: "Batch effect" in MR images, due to vendor-specific features, MR machine generations, and imaging parameters, challenges image quality and hinders deep learning (DL) model generalizability. PURPOSE: We aim to develop a DL model using contrast adjustment and super-resolution to reduce diffusion-weighted images (DWIs) diversity across magnetic field strengths and imaging parameters. STUDY TYPE: Retrospective. SUBJECTS: The DL model was built using an open dataset from one individual. The MR machine identification model was trained and validated on a dataset of 1134 adults (54% females, 46% males), with 1050 subjects showing no DWI abnormalities and 84 with conditions like stroke and tumors. The 21,000 images were divided into 80% for training, 20% for validation, and 3500 for testing. FIELD STRENGTH/SEQUENCE: Seven MR scanners from four manufacturers with 1.5 T and 3 T magnetic field strengths. DWIs were acquired using spin-echo sequences and high-resolution T2WIs using the T2-SPACE sequence. ASSESSMENT: An experienced, board-certified radiologist evaluated the effectiveness of restoring high-resolution T2WI and harmonizing diverse DWI with metrics such as PSNR and SSIM, and the texture and frequency attributes were further analyzed using gray-level co-occurrence matrix and 1-dimensional power spectral density. The model's impact on machine-specific characteristics was gauged through the performance metrics of a ResNet-50 model. Comprehensive statistical tests were employed for statistical robustness, including McNemar's test and the Dice index. RESULTS: Our DL protocol reduced DWI contrast and resolution variation. ResNet-50 model's accuracy decreased from 0.9443 to 0.5786, precision from 0.9442 to 0.6494, recall from 0.9443 to 0.5786, and F1 score from 0.9438 to 0.5587. The t-SNE visualization indicated more consistent image features across multiple MR devices. Autoencoder halved learning iterations; Dice coefficient >0.74 confirmed signal reproducibility in 84 lesions. CONCLUSION: This study presents a DL strategy to mitigate batch effects in diffusion MR images, improving their quality and generalizability. EVIDENCE LEVEL: 3 TECHNICAL EFFICACY: Stage 1.

Glymphatic system impairment in corticobasal syndrome: diffusion tensor image analysis along the perivascular space (DTI-ALPS).

PURPOSE: This study aimed to evaluate the along the perivascular space (ALPS) index based on the diffusion tensor image ALPS (DTI-ALPS) in corticobasal degeneration with corticobasal syndrome (CBD-CBS) and investigate its correlation with motor and cognitive functions. MATERIALS AND METHODS: The data of 21 patients with CBD-CBS and 17 healthy controls (HCs) were obtained from the 4-Repeat Tauopathy Neuroimaging Initiative and the Frontotemporal Lobar Degeneration Neuroimaging Initiative databases. Diffusion magnetic resonance imaging was performed using a 3-Tesla MRI scanner. The ALPS index based on DTI-ALPS was automatically calculated after preprocessing. The ALPS index was compared between the CBD-CBS and HC groups via a general linear model analysis, with covariates such as age, sex, years of education, and intracranial volume (ICV). Furthermore, to confirm the relation between the ALPS index and the motor and cognitive score in CBD-CBS, the partial Spearman's rank correlation coefficient was calculated with covariates such as age, sex, years of education, and ICV. A p value of < 0.05 was considered as statistically significant in all statistical analyses. RESULTS: The ALPS index of CBD-CBS was significantly lower than that of HC (Cohen's d = - 1.53, p < 0.005). Moreover, the ALPS index had a significant positive correlation with the mini mental state evaluation score (rs = 0.65, p < 0.005) and a significant negative correlation with the unified Parkinson's Disease Rating Scale III score (rs = - 0.75, p < 0.001). CONCLUSION: The ALPS index of patients with CBD-CBS, which is significantly lower than that of HCs, is significantly associated with motor and cognitive functions.

Glymphatic system impairment in sleep disruption: diffusion tensor image analysis along the perivascular space (DTI-ALPS).

PURPOSE: This study aimed to evaluate the relationship between sleep quality as assessed using the Pittsburgh Sleep Quality Index (PSQI) and the index of diffusivity along the perivascular space (ALPS index), a possible indirect indicator of glymphatic system activity. MATERIALS AND METHODS: This study included the diffusion magnetic resonance imaging (MRI) data of 317 people with sleep disruption and 515 healthy controls (HCs) from the Human Connectome Project (WU-MINN HCP 1200). The ALPS index was calculated automatically based on diffusion tensor image analysis (DTI)-ALPS of diffusion MRI. The ALPS index of the sleep disruption and HC groups was compared using general linear model (GLM) analysis with covariates, such as age, sex, level of education, and intracranial volume. In addition, to confirm the relationship between sleep quality and the ALPS index in the sleep disruption group as well as evaluate the effect of each PSQI component on the ALPS index, correlation analyses between the ALPS indices and PSQI scores of all the components and between the ALPS index and each PSQI component was performed using GLM analysis with the abovementioned covariates, respectively. RESULTS: The ALPS index was significantly lower in the sleep disruption group than in the HC group (p = 0.001). Moreover, the ALPS indices showed significant negative correlations with the PSQI scores of all the components (false discovery rate [FDR]-corrected p < 0.001). Two significant negative correlations were also found between the ALPS index and PSQI component 2 (sleep latency, FDR-corrected p < 0.001) and 6 (the use of sleep medication, FDR-corrected p < 0.001). CONCLUSION: Our findings suggest that glymphatic system impairment contributes to sleep disruption in young adults.

Reproducibility of automated calculation technique for diffusion tensor image analysis along the perivascular space.

PURPOSE: The method of diffusion tensor image analysis along the perivascular space (DTI-ALPS) was gathering attention to evaluate the brain's glymphatic function or interstitial fluid dynamics. However, to the best knowledge, no study was conducted on the reproducibility of these automated methods for ALPS index values. Therefore, the current study evaluated the ALPS index reproducibility based on DTI-ALPS using two major automated calculation techniques in scan and rescan of the same subject on the same day. MATERIALS AND METHODS: This study included 23 participants, including 2 with Alzheimer's disease, 15 with mild cognitive impairment, and 6 with cognitive normals. Scan and rescan data of diffusion magnetic resonance images were obtained, as well as automatically index for ALPS (ALPS index) and ALPS index maintaining tensor vector orientation information (vALPS index) with region of interest on the template fractional anisotropy map calculated by FSL software.These ALPS indices were compared in terms of scan and rescan reproducibility. RESULTS: The absolute difference in ALPS-index values between scan and rescan was larger in the ALPS index than in the vALPS index by approximately 0.6% as the relative difference. Cohen's d for the left and right ALPS indices between methods were 0.121 and 0.159, respectively. CONCLUSION: The vALPS index based on DTI-ALPS maintaining tensor vector orientation information has higher reproducibility than the ALPS index. This result encourages a multisite study on the ALPS index with a large sample size and helps detect a subtle pathological change in the ALPS index.

Association of MRI Indices of Glymphatic System With Amyloid Deposition and Cognition in Mild Cognitive Impairment and Alzheimer Disease.

BACKGROUND AND OBJECTIVES: The glymphatic system is a whole-brain perivascular network, which promotes CSF/interstitial fluid exchange. Alterations to this system may play a pivotal role in amyloid ß (Aß) accumulation. However, its involvement in Alzheimer's disease (AD) pathogenesis is not fully understood. Here, we investigated the changes in noninvasive MRI measurements related to the perivascular network in patients with mild cognitive impairment (MCI) and AD. Additionally, we explored the associations of MRI measures with neuropsychological score, PET standardized uptake value ratio (SUVR), and Aß deposition. METHODS: MRI measures, including perivascular space (PVS) volume fraction (PVSVF), fractional volume of free water in white matter (FW-WM), and index of diffusivity along the perivascular space (ALPS index) of patients with MCI, those with AD, and healthy controls from the Alzheimer's Disease Neuroimaging Initiative database were compared. MRI measures were also correlated with the levels of CSF biomarkers, PET SUVR, and cognitive score in the combined subcohort of patients with MCI and AD. Statistical analyses were performed with age, sex, years of education, and APOE status as confounding factors. RESULTS: In total, 36 patients with AD, 44 patients with MCI, and 31 healthy controls were analyzed. Patients with AD had significantly higher total, WM, and basal ganglia PVSVF (Cohen's d = 1.15-1.48; p < 0.001), and FW-WM (Cohen's d = 0.73; p < 0.05) and a lower ALPS index (Cohen's d = 0.63; p < 0.05) than healthy controls. Meanwhile, the MCI group only showed significantly higher total (Cohen's d = 0.99; p < 0.05) and WM (Cohen's d = 0.91; p < 0.05) PVSVF. Low ALPS index was associated with lower CSF Aß42 (r s = 0.41, p fdr = 0.026), FDG-PET uptake (r s = 0.54, p fdr < 0.001), and worse multiple cognitive domain deficits. High FW-WM was also associated with lower CSF Aß42 (r s = -0.47, p fdr = 0.021) and worse cognitive performances. CONCLUSION: Our study indicates that changes in PVS-related MRI parameters occur in MCI and AD, possibly due to impairment of the glymphatic system. We also report the associations between MRI parameters and Aß deposition, neuronal change, and cognitive impairment in AD.

Computer-aided diagnosis of lung nodule using gradient tree boosting and Bayesian optimization.

We aimed to evaluate a computer-aided diagnosis (CADx) system for lung nodule classification focussing on (i) usefulness of the conventional CADx system (hand-crafted imaging feature + machine learning algorithm), (ii) comparison between support vector machine (SVM) and gradient tree boosting (XGBoost) as machine learning algorithms, and (iii) effectiveness of parameter optimization using Bayesian optimization and random search. Data on 99 lung nodules (62 lung cancers and 37 benign lung nodules) were included from public databases of CT images. A variant of the local binary pattern was used for calculating a feature vector. SVM or XGBoost was trained using the feature vector and its corresponding label. Tree Parzen Estimator (TPE) was used as Bayesian optimization for parameters of SVM and XGBoost. Random search was done for comparison with TPE. Leave-one-out cross-validation was used for optimizing and evaluating the performance of our CADx system. Performance was evaluated using area under the curve (AUC) of receiver operating characteristic analysis. AUC was calculated 10 times, and its average was obtained. The best averaged AUC of SVM and XGBoost was 0.850 and 0.896, respectively; both were obtained using TPE. XGBoost was generally superior to SVM. Optimal parameters for achieving high AUC were obtained with fewer numbers of trials when using TPE, compared with random search. Bayesian optimization of SVM and XGBoost parameters was more efficient than random search. Based on observer study, AUC values of two board-certified radiologists were 0.898 and 0.822. The results show that diagnostic accuracy of our CADx system was comparable to that of radiologists with respect to classifying lung nodules.

Model-based iterative reconstruction for detection of subtle hypoattenuation in early cerebral infarction: a phantom study.

PURPOSE: Model-based iterative reconstruction (MBIR) was recently shown to enable dose reduction in computed tomography (CT). The detectability of low-contrast lesions was assessed on CT images reconstructed with MBIR compared with the conventional filtered back-projection (FBP) method. MATERIALS AND METHODS: A phantom simulating brain gray matter containing small lesions mimicking early cerebral infarctions was scanned at tube currents of 50, 100, 200, and 400 mA. Images were reconstructed by use of both methods. Round regions were cropped from the reconstructed images, half with a lesion, the other half without. Eight radiologists reviewed the images and scored the certainty of lesion detection on a 5-point scale. Overall performance was analyzed by use of a receiver operating characteristic curve. RESULTS: For the tube currents investigated, the analysis showed that the mean areas under the curves for the reviewers were 0.65, 0.70, 0.82, and 0.83 for FBP and 0.70, 0.76, 0.78, and 0.90 for MBIR. For each current, there was no significant difference between the areas under the curves for the different reconstruction methods (p = 0.32, 0.24, 0.49, and 0.17). CONCLUSION: For the small, low-contrast lesions in the phantom model used in this study, no significant difference between detectability was observed for MBIR and FBP.

Prevalence and diagnostic performance of computed tomography angiography spot sign for intracerebral hematoma expansion depend on scan timing.

INTRODUCTION: The computed tomography angiography (CTA) spot sign correlates with intracerebral hemorrhage (ICH) expansion; however, various diagnostic performances for hematoma expansion, especially in sensitivity, have been reported. We aimed to assess the impact of scan timing of CTA on the diagnostic performance of the CTA spot sign for ICH expansion in two different arterial phases within patients. METHODS: Eighty-three consecutive patients with primary ICH who received two sequential CTAs were recruited. Two neuroradiologists reviewed CTAs for CTA spot signs, while one reviewed initial and follow-up non-contrast CT for measuring ICH volume. The time interval between two phases was then calculated, and the diagnostic performance of CTA spot sign in each phase was evaluated. RESULTS: CTA spot signs were observed in 20/83 (24.1 %) patients in the early phase and 44/83 (53.0%) patients in the late phase. The mean time interval between the two phases was 12.7 s. Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy for hematoma progression of CTA spot sign were 48.1, 87.5, 65.0, 77.8, and 74.7%, respectively, in early phase and 92.6, 66.1, 56.8, 94.9, and 74.7%, respectively, in late phase. The CTA spot sign was significantly associated with ICH expansion in early (P < 0.001) and late (P < 0.00001) phases (Pearson's chi-square test). CONCLUSION: A mere 10-s difference in scan timing could make a difference on prevalence and diagnostic performance of the CTA spot sign, suggesting a need for the standardization of the CTA protocol to generalize the approach for effective clinical application.

Neuromelanin magnetic resonance imaging reveals increased dopaminergic neuron activity in the substantia nigra of patients with schizophrenia.

PURPOSE: The dopamine hypothesis suggests that excessive dopamine release results in the symptoms of schizophrenia. The purpose of this study was to elucidate the dopaminergic and noradrenergic neurons using 3-T neuromelanin magnetic resonance imaging (MRI) in patients with schizophrenia and healthy control subjects. METHODS: We prospectively examined 52 patients with schizophrenia (M: F = 27∶25, mean age, 35 years) and age- and sex-matched healthy controls. Using a 3T MRI unit, we obtained oblique T1-weighted axial images perpendicular to the brainstem. We measured the signal intensity and area for the substantia nigra (SNc), midbrain tegmentum, locus ceruleus (LC), and pons. We then calculated the contrast ratios (CR) for the SNc (CRSN) and LC (CRLC), which were compared between patients and healthy controls using unpaired t-tests. RESULTS: The SNc and LC were readily identified in both patients and healthy controls as areas with high signal intensities in the posterior part of the cerebral peduncle and in the upper pontine tegmentum. The CRSN values in patients were significantly higher than those in healthy controls (10.89±2.37 vs. 9.6±2.36, p<0.01). We observed no difference in the CRLC values between the patients and healthy controls (14.21±3.5 vs. 13.44±3.37, p = 0.25). Furthermore, there was no difference in area of the SNc and LC between schizophrenia patients and controls. CONCLUSIONS: Neuromelanin MRI might reveal increased signal intensity in the SNc of patients with schizophrenia. Our results indicate the presence of excessive dopamine products in the SNc of these patients.

Dual-energy CT for detection of contrast enhancement or leakage within high-density haematomas in patients with intracranial haemorrhage.

INTRODUCTION: Our study aimed to elucidate the diagnostic performance of dual-energy CT (DECT) in the detection of contrast enhancement in intracranial haematomas (ICrH) with early phase dual-energy computed tomography angiography (CTA) and compare the results with those obtained by delayed CT enhancement. METHODS: Thirty-six patients with ICrH were retrospectively included in this study. All patients had undergone single-energy non-contrast CT and contrast-enhanced dual-source DECT. DECT images were post-processed with commercial software, followed by obtaining iodine images and virtual non-contrast images and generating combined images that created the impression of 120-kVp images. Two neuroradiologists, blinded to the patients' data, reviewed two reading sessions: session A (non-contrast CT and combined CT) and session B (non-contrast CT, combined CT, and iodine images) for detection of contrast enhancement in the haematomas. RESULTS: Contrast leakage or enhancement was detected in 23 (57.5 %) out of 40 haemorrhagic lesions in 36 patients on delayed CT. Three enhanced lesions were depicted only in the DECT iodine images. The sensitivity, specificity, positive predictive value, and negative predictive value of session A were 82.6, 94.1, 95.0, and 80.0 %, respectively, and those of session B were 95.7, 94.1, 95.7, and 94.1 %, respectively. CONCLUSION: DECT emphasised the iodine enhancement and facilitated the detection of contrast enhancement or leakage.