FLAIR2: A Combination of FLAIR and T2 for Improved MS Lesion Detection

BACKGROUND AND PURPOSE:FLAIR and double inversion recovery are important MR imaging scans for MS. The suppression of signal from CSF in FLAIR and the additional suppression of WM signal in double inversion recovery improve contrast between lesions, WM and GM, albeit at a reduced SNR. However, whether the acquisition of double inversion recovery is necessary is still debated. Here, we present an approach that allows obtaining CSF-suppressed images with improved contrast between lesions, WM and GM without strongly penalizing SNR.MATERIALS AND METHODS:3D T2-weighted and 3D-FLAIR data acquired from September 2014 to April 2015 in healthy volunteers (23.4 ± 2.4 years of age; female/male ratio, 3:2) and patients (44.1 ± 14.0 years of age; female/male ratio, 4:5) with MS were coregistered and multiplied (FLAIR²). SNR and contrast-to-noise measurements were performed for focal lesions and GM and WM. Furthermore, data from 24 subjects with relapsing-remitting and progressive MS were analyzed retrospectively (52.7 ± 8.1 years of age; female/male ratio, 14:10).RESULTS:The GM-WM contrast-to-noise ratio was by 133% higher in FLAIR² than in FLAIR and improved between lesions and WM by 31%, 93%, and 158% compared with T2, DIR, and FLAIR, respectively. Cortical and juxtacortical lesions were more conspicuous in FLAIR². Furthermore, the 3D nature of FLAIR² allowed reliable visualization of callosal and infratentorial lesions.CONCLUSIONS:We present a simple approach for obtaining CSF suppression with an improved contrast-to-noise ratio compared with conventional FLAIR and double inversion recovery without the acquisition of additional data. FLAIR² can be computed retrospectively if T2 and FLAIR scans are available.

MR imaging is important for the diagnosis and monitoring of MS. Formation of MS lesions creates a hydrophilic environment, resulting in an increase in the T2 and proton density–weighted MR signal and a signal reduction on T1-weighted scans.¹ Ovoid hyperintense areas on T2-weighted MR imaging are therefore a radiologic hallmark of MS. Lesion conspicuity is often affected by the bright CSF signal, for instance, close to the ventricles or cortical sulci. FLAIR is a T2-weighted scan that suppresses CSF selectively with an inversion pulse.² Yet, the CSF signal suppression comes at the cost of reduced SNR. Usually, FLAIR scans are acquired in 2D with sections parallel to the subcallosal line. Additional sagittal FLAIR scans are required to reliably detect corpus callosum lesions.^2,3 Furthermore, 2D-FLAIR has artifacts due to CSF and blood inflow and often provides insufficient T2-weighting,⁴ requiring additional proton density/T2-weighted images for the detection of lesions in infratentorial areas. The brain MR imaging protocol for MS studies⁵ includes proton density and T2-weighted spin-echo, axial, and sagittal FLAIR and recommends pre- and postcontrast T1-weighted spin-echo MR imaging.Apart from diagnosis, conventional MR images play an important role as outcome measures in clinical trials of new MS therapies.^5,6 New lesion activity (eg, gadolinium-enhancing lesions and new or enlarging T2-lesions) and estimates of disease burden (eg, total T2-lesion volume or count; T1-hypointense lesion volume; brain atrophy) are typical imaging end points in clinical trials.⁵ These scans are directed toward lesion identification in WM. Demyelination and the appearance of lesions is, however, not limited to the WM; it also involves the deep and cortical GM.⁷ Focal GM lesions appear in the earliest stages of MS^8,9 and are associated with physical and cognitive disability.^10,11 Moreover, cortical lesion load was shown to be a predictor of progression of clinical disability during 5 years¹² and to improve predictions for the conversion from relapsing-remitting to secondary-progressive MS compared with assessing WM lesions alone.¹³ Given the importance of cortical lesions in MS, there is great interest in their visualization. However, the cortex is thin, its myelin content is low, and inflammation is low in cortical lesions. Contrast between lesions and healthy tissue is therefore low, making the detection of cortical damage challenging.In double inversion recovery (DIR),¹⁴ both CSF and the WM signal are suppressed; this suppression results in enhanced contrast between lesions, GM and WM. T1-relaxation times of GM and WM are similar. Therefore, both tissues are affected by the inversion pulse, resulting in reduced SNR. Long data-acquisition times further limit the spatial resolution of DIR to 1 mm³ isotropic at 3T. In a postmortem study, the specificity of 3D-DIR was found to be 90%, whereas sensitivity was only 18%.¹⁵ DIR detected most leukocortical lesions; however, intracortical and subpial lesions were still missed.¹⁵ Intracortical and subpial lesions are the most common cortical lesions in patients with chronic MS, yet subpial lesions are rarely detected with DIR or other techniques.^16,17 More recently, 3D versions of MR imaging sequences for MS have become available¹⁸ but are not yet used widely in clinical imaging of MS.¹⁹ 3D sequences with isotropic voxels of 1 mm³ volume or smaller are particularly suitable for the assessment of the cortex. Moreover, these scans allow simultaneous assessment of all 3 orthogonal image planes. A drawback is the increased acquisition time per scan, in particular for DIR. Lesion detection, especially within the cortex, would benefit from a rapid 3D imaging approach with high spatial resolution, suppressed CSF, higher SNR than DIR, and a good contrast-to-noise ratio (CNR) between lesions, GM and WM.This study aims to develop and test a method that combines the good SNR of T2-weighted images with the CSF suppression of FLAIR to achieve GM-WM contrast similar to that in DIR and good contrast between lesions, healthy tissue. We compared SNR and CNR of this new approach with conventional FLAIR, T2, and DIR; and we present images acquired in patients with relapsing-remitting and progressive MS.     

Detection of lesions in multiple sclerosis by 2D FLAIR and single-slab 3D FLAIR sequences at 3.0 T: initial results

The aim of this study was to compare conventional 2D FLAIR and single-slab 3D FLAIR sequences in the detection of lesions in patients with multiple sclerosis. Eight patients with MS were examined at 3.0 T by using a 2D FLAIR sequence and a single-slab 3D FLAIR sequence. A comparison of lesion detectability was performed for the following regions: periventricular, nonperiventricular/juxtacortical and infratentorial. The contrast-to-noise ratios (CNRs) between lesions and brain tissue and CSF were calculated for each sequence. A total of 424 lesions were found using the 2D FLAIR sequence, while with the 3D FLAIR sequence 719 lesions were found. With the 2D FLAIR sequence, 41% fewer lesions were detected than with the 3D FLAIR sequence. Further, 40% fewer supratentorial and 62.5% fewer infratentorial lesions were found with the 2D FLAIR sequence. In images acquired with the 3D FLAIR sequence, the lesions had significantly higher CNRs than in images acquired with the 2D FLAIR sequence. These are the first results using a single-slab 3D FLAIR sequence at 3.0 T for detection of lesions in patients with MS. With the 3D FLAIR sequence significantly higher CNRs were achieved and significantly more lesions in patients with MS were detected.     

A Novel Deep Learning Approach with a 3D Convolutional Ladder Network for Differential Diagnosis of Idiopathic Normal Pressure Hydrocephalus and Alzheimer’s Disease

Purpose:Idiopathic normal pressure hydrocephalus (iNPH) and Alzheimer’s disease (AD) are geriatric diseases and common causes of dementia. Recently, many studies on the segmentation, disease detection, or classification of MRI using deep learning have been conducted. The aim of this study was to differentiate iNPH and AD using a residual extraction approach in the deep learning method.Methods:Twenty-three patients with iNPH, 23 patients with AD and 23 healthy controls were included in this study. All patients and volunteers underwent brain MRI with a 3T unit, and we used only whole-brain three-dimensional (3D) T₁-weighted images. We designed a fully automated, end-to-end 3D deep learning classifier to differentiate iNPH, AD and control. We evaluated the performance of our model using a leave-one-out cross-validation test. We also evaluated the validity of the result by visualizing important areas in the process of differentiating AD and iNPH on the original input image using the Gradient-weighted Class Activation Mapping (Grad-CAM) technique.Results:Twenty-one out of 23 iNPH cases, 19 out of 23 AD cases and 22 out of 23 controls were correctly diagnosed. The accuracy was 0.90. In the Grad-CAM heat map, brain parenchyma surrounding the lateral ventricle was highlighted in about half of the iNPH cases that were successfully diagnosed. The medial temporal lobe or inferior horn of the lateral ventricle was highlighted in many successfully diagnosed cases of AD. About half of the successful cases showed nonspecific heat maps.Conclusion:Residual extraction approach in a deep learning method achieved a high accuracy for the differential diagnosis of iNPH, AD, and healthy controls trained with a small number of cases.     

A Three-Dimensional Deep Convolutional Neural Network for Automatic Segmentation and Diameter Measurement of Type B Aortic Dissection

ObjectiveTo provide an automatic method for segmentation and diameter measurement of type B aortic dissection (TBAD).Materials and MethodsAortic computed tomography angiographic images from 139 patients with TBAD were consecutively collected. We implemented a deep learning method based on a three-dimensional (3D) deep convolutional neural (CNN) network, which realizes automatic segmentation and measurement of the entire aorta (EA), true lumen (TL), and false lumen (FL). The accuracy, stability, and measurement time were compared between deep learning and manual methods. The intra- and inter-observer reproducibility of the manual method was also evaluated.ResultsThe mean dice coefficient scores were 0.958, 0.961, and 0.932 for EA, TL, and FL, respectively. There was a linear relationship between the reference standard and measurement by the manual and deep learning method (r = 0.964 and 0.991, respectively). The average measurement error of the deep learning method was less than that of the manual method (EA, 1.64% vs. 4.13%; TL, 2.46% vs. 11.67%; FL, 2.50% vs. 8.02%). Bland-Altman plots revealed that the deviations of the diameters between the deep learning method and the reference standard were −0.042 mm (−3.412 to 3.330 mm), −0.376 mm (−3.328 to 2.577 mm), and 0.026 mm (−3.040 to 3.092 mm) for EA, TL, and FL, respectively. For the manual method, the corresponding deviations were −0.166 mm (−1.419 to 1.086 mm), −0.050 mm (−0.970 to 1.070 mm), and −0.085 mm (−1.010 to 0.084 mm). Intra- and inter-observer differences were found in measurements with the manual method, but not with the deep learning method. The measurement time with the deep learning method was markedly shorter than with the manual method (21.7 ± 1.1 vs. 82.5 ± 16.1 minutes, p < 0.001).ConclusionThe performance of efficient segmentation and diameter measurement of TBADs based on the 3D deep CNN was both accurate and stable. This method is promising for evaluating aortic morphology automatically and alleviating the workload of radiologists in the near future.     

Development of Convolutional Neural Networks to identify bone metastasis for prostate cancer patients in bone scintigraphy

Annals of Nuclear Medicine - The main aim of this work is to build a robust Convolutional Neural Network (CNN) algorithm that efficiently and quickly classifies bone scintigraphy images, by...     

Improved Lesion Conspicuity with Contrast-Enhanced 3D T1 TSE Black-Blood Imaging in Cranial Neuritis: A Comparative Study of Contrast-Enhanced 3D T1 TSE, 3D T1 Fast-Spoiled Gradient Echo,and 3D T2 FLAIR

BACKGROUND AND PURPOSE:Contrast-enhanced 3D-turbo spin-echo (TSE) black-blood sequence has gained attention, as it suppresses signals from vessels and provides an increased contrast-noise ratio. The purpose was to investigate which among the contrast-enhanced 3D T1 TSE, 3D T1 fast-spoiled gradient echo (FSPGR), and 3D T2 FLAIR sequences can better detect cranial nerve contrast enhancement.MATERIALS AND METHODS:Patients with cranial neuritis based on clinical findings (n = 20) and control participants (n = 20) were retrospectively included in this study. All patients underwent 3T MR imaging with contrast-enhanced 3D T1 TSE, 3D T1 FSPGR, and 3D T2 FLAIR. Experienced and inexperienced reviewers independently evaluated the 3 sequences to compare their diagnostic performance and time required to reach the diagnosis. Additionally, tube phantoms containing varying concentrations of gadobutrol solution were scanned using the 3 sequences.RESULTS:For the inexperienced reader, the 3D T1 TSE sequence showed significantly higher sensitivity (80% versus 50%, P = .049; 80% versus 55%; P = .040), specificity (100% versus 65%, P = .004; 100% versus 60%; P = .001), and accuracy (90% versus 57.5%, P = .001; 90% versus 57.5%, P = .001) than the 3D T1 FSPGR and 3D T2 FLAIR sequences in patients with cranial neuritis. For the experienced reader, the 3D T1-based sequences showed significantly higher sensitivity than the 3D T2 FLAIR sequence (85% versus 30%, P < .001; 3D T1 TSE versus 3D T2 FLAIR, 85% versus 30%, P < .001; 3D T1 FSPGR versus 3D T2 FLAIR). For both readers, the 3D T1 TSE sequence showed the highest area under the curve (inexperienced reader; 0.91, experienced reader; 0.87), and time to diagnosis was significantly shorter with 3D T1 TSE than with 3D T1 FSPGR.CONCLUSIONS:The 3D T1 TSE sequence may be clinically useful in evaluating abnormal cranial nerve enhancement, especially for inexperienced readers.

Cranial neuropathies can have multiple causes, including infectious, neoplastic, inflammatory, traumatic, and idiopathic pathologies.¹ Such conditions cause disruption of the blood–nerve barrier, which is sustained by the combined actions of tight junctions in the endothelium of the endoneurial capillaries and of the inner layers of the perineurium.² Contrast-enhanced (CE) MR imaging plays an important role in the diagnosis of cranial neuritis by visualizing nerve enhancement attributed to leakage forcing spillage and accumulation of contrast material surrounded by CSF.³To date, no standard protocol has been established for evaluating cranial nerve enhancement, whereas several sequences have been proposed for detecting leptomeningeal enhancement. CE 3D T1 gradient-echo (GRE) sequences have been widely used in the clinical setting to detect leptomeningeal pathology.^4-7 Furthermore, the CE 3D FLAIR sequence is advantageous because it can sensitively detect low concentrations of gadolinium.^8,9 Recently, a CE 3D turbo spin-echo (TSE) black-blood sequence has gained attention because it provides an increased contrast to noise ratio (CNR) and suppresses diverting signals from vessels.^10-13To the best of our knowledge, no study has explored the value of CE 3D T1 TSE black-blood imaging in the diagnosis of cranial neuritis. Although the CE 3D T1 GRE sequence is generally used for the evaluation of cranial nerve enhancement,^3,14 its ability to evaluate the cisternal segment of cranial nerves is limited owing to the surrounding prominent vessel enhancement. Moreover, hyperintensities on FLAIR are also associated with various conditions, such as subarachnoid hemorrhage, sluggish collateral vessels, and supplemental oxygen, which may produce misinterpretations of the cranial nerve enhancement.¹⁵ Therefore, the aim of this study was to investigate which sequence among 3D T1 TSE, 3D T1 fast-spoiled gradient echo (FSPGR), and 3D T2 FLAIR can better detect contrast enhancement in patients with cranial neuritis.     

Effectiveness of 3D T2-Weighted FLAIR FSE Sequences with Fat Suppression for Detection of Brain MR Imaging Signal Changes in Children

BACKGROUND AND PURPOSE:T2-weighted FLAIR can be combined with 3D-FSE sequences with isotropic voxels, yielding higher signal-to-noise ratio than 2D-FLAIR. Our aim was to explore whether a T2-weighted FLAIR–volume isotropic turbo spin-echo acquisition sequence (FLAIR-VISTA) with fat suppression shows areas of abnormal brain T2 hyperintensities with better conspicuity in children than a single 2D-FLAIR sequence.MATERIALS AND METHODS:One week after a joint training session with 20 3T MR imaging examinations (8 under sedation), 3 radiologists independently evaluated the presence and conspicuity of abnormal areas of T2 hyperintensities of the brain in FLAIR-VISTA with fat suppression (sagittal source and axial and coronal reformatted images) and in axial 2D-FLAIR without fat suppression in a test set of 100 3T MR imaging examinations (34 under sedation) of patients 2–18 years of age performed for several clinical indications. Their agreement was measured with weighted κ statistics.RESULTS:Agreement was “substantial” (mean, 0.61 for 3 observers; range, 0.49–0.69 for observer pairs) for the presence of abnormal T2 hyperintensities and “fair” (mean, 0.29; range, 0.23–0.38) for the comparative evaluation of lesion conspicuity. In 21 of 23 examinations in which the 3 radiologists agreed on the presence of abnormal T2 hyperintensities, FLAIR-VISTA with fat suppression images were judged to show hyperintensities with better conspicuity than 2D-FLAIR. In 2 cases, conspicuity was equal, and in no case was conspicuity better in 2D-FLAIR.CONCLUSIONS:FLAIR-VISTA with fat suppression can replace the 2D-FLAIR sequence in brain MR imaging protocols for children.

3D (volume) gradient-echo T1-weighted sequences are a well-established part of brain MR imaging protocols due to the intrinsically higher SNR compared with 2D sequences and the ability to obtain optimal MPR.¹ However, abnormalities of the brain are usually detected as nonspecific areas of variably increased signal in T2WI. FLAIR images are preferable to FSE images for detecting such T2 abnormalities because suppression of the CSF high signal results in an improved gray-scale dynamic range.²T2-weighted FLAIR can be combined with 3D-FSE sequences with isotropic voxels that are variably named by different vendors, including volume isotropic turbo spin-echo acquisition (VISTA; Philips Healthcare, Best, the Netherlands), SPACE (sampling perfection with application-optimized contrasts by using different flip angle evolution; Siemens, Erlangen, Germany), Cube (GE Healthcare, Milwaukee, Wisconsin), isoFSE (http://www.hitachimed.com/products/mri/oasis/Neurological/isoFSE), and 3D mVox (Toshiba, Tokyo, Japan). Such T2-weighted FLAIR 3D-FSE sequences have a higher SNR than 2D-FLAIR, enable MPR, and are less affected by CSF flow artifacts,^3–6 which are more prominent in sedated children at a higher field strength 3T magnet.^7–9Theoretically, suppression of fat signal with spectral presaturation could improve the sensitivity of FLAIR-VISTA by further narrowing the gray-scale dynamic range.²The purpose of the present study was to evaluate whether a FLAIR-VISTA sequence with fat suppression shows abnormal brain T2 signal hyperintensities with better conspicuity than a 2D-FLAIR sequence on a single axial plane in children.     

Deep Neural Networks Predict the Need for CT in Pediatric Mild Traumatic Brain Injury: A Corroboration of the PECARN Rule

PurposeOnly 10% of CT scans unveil positive findings in mild traumatic brain injury, raising concerns of its overuse in this population. A number of clinical rules have been developed to address this issue, but they still suffer limitations in their specificity. Machine learning models have been applied in limited studies to mimic clinical rules; however, further improvement in terms of balanced sensitivity and specificity is still needed. In this work, the authors applied a deep artificial neural networks (DANN) model and an instance hardness threshold algorithm to reproduce the Pediatric Emergency Care Applied Research Network (PECARN) clinical rule in a pediatric population collected as a part of the PECARN study between 2004 and 2006.MethodsThe DANN model was applied using 14,983 patients younger than 18 years with Glasgow Coma Scale scores ≥ 14 who had head CT reports. The clinical features of the PECARN rules, PECARN-A (group A, age < 2 years) and PECARN-B (group B, age ≤ 2 years), were used to directly evaluate the model. The average accuracy, sensitivity, precision, and specificity were calculated by comparing the model’s prediction outcome to that reported by the PECARN investigators. The instance hardness threshold and DANN model were applied to predict the need for CT in pediatric patients using 5-fold cross-validation.ResultsIn the first phase, the DANN model resulted in 98.6% sensitivity and 99.7% specificity for predicting the need for CT using the predictors of the two PECARN clinical rules combined to train the model. In the second phase, the DANN model was superior to both the PECARN-A and PECARN-B rules using the predictors for each age group separately to train the model. Compared with the clinical rule, for group A, the model achieved an average sensitivity (93.7% versus 100%) and specificity (97.5% versus 53.6%); for group B, the average sensitivity of the model was 99.2% versus 98.6%, and the specificity was 98.8% versus 58.2%.ConclusionsIn this study, a DANN model achieved comparable sensitivity and outstanding specificity for replicating the PECARN clinical rule and predicting the need for CT in pediatric patients after mild traumatic brain injury compared with the original statistically derived clinical rule.     

Deep Learning Algorithm for Simultaneous Noise Reduction and Edge Sharpening in Low-Dose CT Images: A Pilot Study Using Lumbar Spine CT

ObjectiveThe purpose of this study was to assess whether a deep learning (DL) algorithm could enable simultaneous noise reduction and edge sharpening in low-dose lumbar spine CT.Materials and MethodsThis retrospective study included 52 patients (26 male and 26 female; median age, 60.5 years) who had undergone CT-guided lumbar bone biopsy between October 2015 and April 2020. Initial 100-mAs survey images and 50-mAs intraprocedural images were reconstructed by filtered back projection. Denoising was performed using a vendor-agnostic DL model (ClariCT.AI™, ClariPI) for the 50-mAS images, and the 50-mAs, denoised 50-mAs, and 100-mAs CT images were compared. Noise, signal-to-noise ratio (SNR), and edge rise distance (ERD) for image sharpness were measured. The data were summarized as the mean ± standard deviation for these parameters. Two musculoskeletal radiologists assessed the visibility of the normal anatomical structures.ResultsNoise was lower in the denoised 50-mAs images (36.38 ± 7.03 Hounsfield unit [HU]) than the 50-mAs (93.33 ± 25.36 HU) and 100-mAs (63.33 ± 16.09 HU) images (p < 0.001). The SNRs for the images in descending order were as follows: denoised 50-mAs (1.46 ± 0.54), 100-mAs (0.99 ± 0.34), and 50-mAs (0.58 ± 0.18) images (p < 0.001). The denoised 50-mAs images had better edge sharpness than the 100-mAs images at the vertebral body (ERD; 0.94 ± 0.2 mm vs. 1.05 ± 0.24 mm, p = 0.036) and the psoas (ERD; 0.42 ± 0.09 mm vs. 0.50 ± 0.12 mm, p = 0.002). The denoised 50-mAs images significantly improved the visualization of the normal anatomical structures (p < 0.001).ConclusionDL-based reconstruction may enable simultaneous noise reduction and improvement in image quality with the preservation of edge sharpness on low-dose lumbar spine CT. Investigations on further radiation dose reduction and the clinical applicability of this technique are warranted.     

Utility of a Multiparametric Quantitative MRI Model That Assesses Myelin and Edema for Evaluating Plaques,Periplaque White Matter,and Normal-Appearing White Matter in Patients with Multiple Sclerosis: A Feasibility Study

BACKGROUND AND PURPOSE:T1 and T2 values and proton density can now be quantified on the basis of a single MR acquisition. The myelin and edema in a voxel can also be estimated from these values. The purpose of this study was to evaluate a multiparametric quantitative MR imaging model that assesses myelin and edema for characterizing plaques, periplaque white matter, and normal-appearing white matter in patients with MS.MATERIALS AND METHODS:We examined 3T quantitative MR imaging data from 21 patients with MS. The myelin partial volume, excess parenchymal water partial volume, the inverse of T1 and transverse T2 relaxation times (R1, R2), and proton density were compared among plaques, periplaque white matter, and normal-appearing white matter.RESULTS:All metrics differed significantly across the 3 groups (P < .001). Those in plaques differed most from those in normal-appearing white matter. The percentage changes of the metrics in plaques and periplaque white matter relative to normal-appearing white matter were significantly more different from zero for myelin partial volume (mean, −61.59 ± 20.28% [plaque relative to normal-appearing white matter], and mean, −10.51 ± 11.41% [periplaque white matter relative to normal-appearing white matter]), and excess parenchymal water partial volume (13.82 × 10³ ± 49.47 × 10³% and 51.33 × 10² ± 155.31 × 10²%) than for R1 (−35.23 ± 13.93% and −6.08 ± 8.66%), R2 (−21.06 ± 11.39% and −4.79 ± 6.79%), and proton density (23.37 ± 10.30% and 3.37 ± 4.24%).CONCLUSIONS:Multiparametric quantitative MR imaging captures white matter damage in MS. Myelin partial volume and excess parenchymal water partial volume are more sensitive to the MS disease process than R1, R2, and proton density.

MS is an inflammatory demyelinating disorder of the central nervous system, which mainly affects young adults. MR imaging plays a major role in the diagnosis and surveillance of patients with MS for initial and follow-up detection of focal cerebral lesions.¹ In addition to conventional MR imaging techniques including T2-weighted imaging, quantitative MR imaging techniques enable characterization of MS lesions and detection of otherwise hidden abnormalities in normal-appearing white matter (NAWM).^2,3 Moreover, diffusion tensor imaging and q-space imaging reveal abnormalities of white matter at the periphery of visible plaques on conventional MR images (periplaque white matter [PWM]) and NAWM^4,5: The fractional anisotropy and apparent diffusion coefficient measured by diffusion tensor imaging and root mean square displacement measured by q-space imaging were worst in plaques, and in PWM, worse than in NAWM.A recently developed MR imaging quantification pulse sequence, QRAPMASTER (quantification of relaxation times and proton density by multiecho acquisition of a saturation-recovery using turbo spin-echo readout), has made it possible to quantify longitudinal T1 and transverse T2 relaxation times, their inverses R1 and R2, and proton density (PD) in a single acquisition in a clinically acceptable time.⁶ A phantom study has shown that these measurements are sufficiently accurate and reproducible for use in clinical practice.⁷ Several studies have also shown the validity of this sequence in evaluating diseases such as metastatic brain tumors⁸ and Sturge-Weber syndrome,^9,10 in addition to MS.¹¹ Even though the synthetic FLAIR image has lower image quality than the conventional FLAIR image, synthetic MR imaging overall has been shown to have comparable diagnostic power with conventional MR imaging for MS, while additionally offering fast and robust volumetry.¹¹ By using the QRAPMASTER pulse sequence, the R1, R2, and PD values of plaques, NAWM, and diffusely abnormal white matter of patients with MS were shown to be different from those of white matter in healthy controls.¹² Furthermore, the myelin partial volume (V_MY) and excess parenchymal water partial volume (V_EPW) can now be estimated from R1, R2, and PD,¹³ to indicate the quantities of myelin and edema, respectively, in the brain. In the pathologic brain, a decrease in V_MY indicating myelin loss or an increase in V_EPW indicating edema will occur in this model. This model postulates 4 partial volume compartments: V_MY, the cellular partial volume (V_CL), the free water partial volume (V_FW), and V_EPW. The model assumes that each compartment has its own R1, R2, and PD values and that the relaxation behaviors of all 4 compartments contribute to the effective R1, R2, and PD values of an acquisition voxel. V_MY contains myelin water and myelin macromolecules. V_CL contains intracellular water, extracellular water, and nonmyelin macromolecules. Myelin water is trapped between myelin sheaths and has a much shorter T2 relaxation time than intracellular and extracellular water. The commonly calculated myelin water fraction corresponds to PD in the V_MY. The proportional relation between myelin water fraction and myelin content has been validated by histopathology.^14,15 V_EPW, or edema, adds water to the V_CL. Because no distinction can be made between excess water and the already present water in the V_CL, the magnetization exchange rate between V_EPW and V_CL is assumed to be infinitely high. V_MY and V_EPW may reflect the disease burden of patients with MS more specifically than R1, R2, and PD.The aim of this study was to evaluate this multiparametric quantitative MR imaging model that assesses myelin and edema for characterizing plaques, PWM, and NAWM in patients with MS.     

Development and Validation of a Deep Learning–Based Automatic Brain Segmentation and Classification Algorithm for Alzheimer Disease Using 3D T1-Weighted Volumetric Images

BACKGROUND AND PURPOSE:Limited evidence has suggested that a deep learning automatic brain segmentation and classification method, based on T1-weighted brain MR images, can predict Alzheimer disease. Our aim was to develop and validate a deep learning–based automatic brain segmentation and classification algorithm for the diagnosis of Alzheimer disease using 3D T1-weighted brain MR images.MATERIALS AND METHODS:A deep learning–based algorithm was developed using a dataset of T1-weighted brain MR images in consecutive patients with Alzheimer disease and mild cognitive impairment. We developed a 2-step algorithm using a convolutional neural network to perform brain parcellation followed by 3 classifier techniques including XGBoost for disease prediction. All classification experiments were performed using 5-fold cross-validation. The diagnostic performance of the XGBoost method was compared with logistic regression and a linear Support Vector Machine by calculating their areas under the curve for differentiating Alzheimer disease from mild cognitive impairment and mild cognitive impairment from healthy controls.RESULTS:In a total of 4 datasets, 1099, 212, 711, and 705 eligible patients were included. Compared with the linear Support Vector Machine and logistic regression, XGBoost significantly improved the prediction of Alzheimer disease (P < .001). In terms of differentiating Alzheimer disease from mild cognitive impairment, the 3 algorithms resulted in areas under the curve of 0.758–0.825. XGBoost had a sensitivity of 68% and a specificity of 70%. In terms of differentiating mild cognitive impairment from the healthy control group, the 3 algorithms resulted in areas under the curve of 0.668–0.870. XGBoost had a sensitivity of 79% and a specificity of 80%.CONCLUSIONS:The deep learning–based automatic brain segmentation and classification algorithm allowed an accurate diagnosis of Alzheimer disease using T1-weighted brain MR images. The widespread availability of T1-weighted brain MR imaging suggests that this algorithm is a promising and widely applicable method for predicting Alzheimer disease.

Alzheimer disease (AD) is the most common cause of dementia, with mild cognitive impairment (MCI) regarded as a transitional state between normal cognition and early stages of dementia.¹ Although current therapeutic and preventive options are only moderately effective, a reliable decision-making diagnostic approach is important during early stages of AD.^2,3 The guidelines of the National Institute on Aging–Alzheimer’s Association suggest that MR imaging is a supportive imaging tool in the diagnostic work-up of patients with AD and MCI.^2,3 Imaging biomarkers play an important role in the diagnosis of AD, both in the research field and in clinical practice. The identification of amyloid and the τ PET ligand provided huge advances in understanding the pathophysiologic mechanisms underlying AD and its early diagnosis, even in the preclinical or prodromal stage.^4-6 Although amyloid and τ PET are more sensitive and specific for the diagnosis of AD, they are expensive to perform, have limited availability, and require ionizing radiation, limiting their use in clinical practice. CSF amyloid and τ are also important biomarkers that could be used for AD diagnostics in the clinical research setting.^3,7-9 However, CSF AD biomarkers also have limited availability. MR imaging, however, is widely available and used in standard practice to support the diagnosis of AD and to exclude other causes of cognitive impairment, including stroke, vascular dementia, normal-pressure hydrocephalus, and inflammatory and neoplastic conditions.3D T1-weighted volumetric MR imaging is the most important MR imaging tool in the diagnosis of AD. 3D volumetry has long been used as a morphologic diagnostic tool for AD, not only as a visual assessment or manual segmentation but for semiautomatic and automatic segmentation. Examples include semiautomatic structural changes on MR imaging,¹⁰ automated hippocampal volumetry,¹¹ entorhinal cortex atrophy,¹² and changes in pineal gland volume.¹³ Although user-friendly automated segmentation algorithms were first introduced 20 years ago, evidence supporting the use of 3D volumetry in clinical practice is currently insufficient. Visual assessment requires experience, and automatic 3D volumetry requires a long acquisition time.To our knowledge, limited evidence has suggested that a deep learning automatic brain segmentation and classification method, based on T1-weighted brain MR images, can predict AD.¹⁴ Currently available algorithms have low clinical feasibility because of the long processing time for brain segmentation, and the classification algorithm based on T1-weighted brain MR images needs to be validated in a large external dataset. The purpose of this study was to develop and validate a deep learning–based automatic brain segmentation and classification algorithm for the diagnosis of AD using 3D T1-weighted brain MR images.     

Deep Learning Network for Segmentation of the Prostate Gland With Median Lobe Enlargement in T2-weighted MR Images: Comparison With Manual Segmentation Method

PurposeAim of this study was to evaluate a fully automated deep learning network named Efficient Neural Network (ENet) for segmentation of prostate gland with median lobe enlargement compared to manual segmentation.Materials and MethodsOne-hundred-three patients with median lobe enlargement on prostate MRI were retrospectively included. Ellipsoid formula, manual segmentation and automatic segmentation were used for prostate volume estimation using T2 weighted MRI images. ENet was used for automatic segmentation; it is a deep learning network developed for fast inference and high accuracy in augmented reality and automotive scenarios. Student t-test was performed to compare prostate volumes obtained with ellipsoid formula, manual segmentation, and automated segmentation. To provide an evaluation of the similarity or difference to manual segmentation, sensitivity, positive predictive value (PPV), dice similarity coefficient (DSC), volume overlap error (VOE), and volumetric difference (VD) were calculated.ResultsDifferences between prostate volume obtained from ellipsoid formula versus manual segmentation and versus automatic segmentation were statistically significant (P < 0.049318 and P < 0.034305, respectively), while no statistical difference was found between volume obtained from manual versus automatic segmentation (P = 0.438045). The performance of ENet versus manual segmentations was good providing a sensitivity of 93.51%, a PPV of 87.93%, a DSC of 90.38%, a VOE of 17.32% and a VD of 6.85%.ConclusionThe presence of median lobe enlargement may lead to MRI volume overestimation when using the ellipsoid formula so that a segmentation method is recommended. ENet volume estimation showed great accuracy in evaluation of prostate volume similar to that of manual segmentation.     

A 3D T1-weighted gradient-echo sequence for routine use in 3D radiosurgical treatment planning of brain metastases: first clinical results

The authors report on a 3D sequence for MRI of the brain and its application in radiosurgical treatment planning of 35 brain metastases. The measuring sequence, called magnetization — prepared rapid gradient echo (MPRAGE), was compared with 2D T1-weighted spin-echo (SE) sequences following intravenous contrast-medium application in 19 patients with brain metastases. The average diameter of all lesions was similar in both sequences, with 16.8 and 17.0 mm for SE and MPRAGE, respectively. Target point definition was equal in 29 metastases, and in 6 cases superior on MPRAGE, due to better gray-white matter contrast and increased contrast enhancement. In cases of bleeding metastases there was improved depiction of internal structures in 3D MRI. Postprocessing of 3D MPRAGE data created multi-planar reconstruction along any chosen plane with isotropic spatial resolution, which helped to improve radiosurgical isodose distribution in 4 cases when compared to 2D SE. However, sensitivity of 3D MPRAGE to detect small lesions (< 3 mm) was decreased in one patient with more than 50 metastases. We conclude that 3D gradient-echo (GE) imaging might be of great value for radiosurgical treatment planning, but does not replace 2D SE with its current parameters. Correspondence to: H. Hawighorst