Magnetic Resonance Imaging Slices Research Articles

Rubik-Net: Learning Spatial Information via Rotation-Driven Convolutions for Brain Segmentation.

The accurate segmentation of brain tissue in Magnetic Resonance Image (MRI) slices is essential for assessing neurological conditions and brain diseases. However, it is challenging to segment MRI slices because of the low contrast between different brain tissues and the partial volume effect. 2-Dimensional (2-D) convolutional networks cannot handle such volumetric image data well because they overlook spatial information between MRI slices. Although 3-Dimensional (3-D) convolutions capture volumetric spatial information, they have not been fully exploited to enhance representative ability of deep networks; moreover, they may lead to overfitting with insufficient training data. In this paper, we propose a novel convolutional mechanism, termed Rubik convolution, to capture multi-dimensional information between MRI slices. Rubik convolution rotates the axis of a set of consecutive slices, enabling 2-D convolution kernels to extract features of each axial plane simultaneously. Next, feature maps are rotated back to fuse multidimensional information by the Max-View-Maps. Furthermore, we propose an efficient 2-D convolutional network, namely Rubik-Net, where the residual connections and the bottleneck structure are used to enhance information transmission and reduce the number of network parameters. The Rubik-Net shows promising results on iSeg2017, iSeg2019, IBSR and BrainWeb datasets in terms of segmentation accuracy. In particular, we achieved the best results in 95th percentile Hausdorff distance and average surface distance in cerebrospinal fluid segmentation on the most challenging iSeg2019 dataset. The experiments indicate that Rubik-Net improves the accuracy and efficiency of medical image segmentation. Moreover, Rubik convolution can be easily embedded into existing 2-D convolutional networks.

Radiological spectrum of temporal bone, inner ear and brain parenchymal abnormalities in the preoperative cross-sectional imaging of sensorineural hearing loss in cochlear implant candidates: A prospective study

Cochlear implantation is a prudent method of rehabilitation in severe sensorineural hearing loss where other means like hearing aids fail. Not everybody with severe sensorineural hearing loss benefits from implantation, in fact is contraindicated in some. A thorough pre-surgical imaging evaluation of inner ear, temporal bone and cerebral parenchyma is important in decision making, more so as the implants and costs involved are quite expensive. Limited studies exist in the Indian setting. Aim was to evaluate the role of combined computed tomography (CT) and Magnetic Resonance imaging (MRI) in pre operative imaging assessment for cochlear implantation, a prospective study was concluded at the Department of Radiodiagnosis over 30 months (2017 to 2019). Patients with severe hearing loss underwent HRCT temporal bone on 128 slice MDCT and MRI with BFFE, T2/Flair etc. Sequences targeting inner ear, cochlear nerve and brain on 3 Tesla scanner. Data was entered in MS EXCEL and analyzed in SPSS software using cross tabulations or correlation coefficient to compare the two imaging modalities. Chi square and P values were obtained for the cross tabulation data of comparison of CT and MRI in assessment of inner ear structures. The study focused on cochleovestibular malformations, patency of the cochlea-vestibule, cochlear nerve status on MRI and to see if there was any correlation with the bony cochlear nerve canal diameter on CT, isolated or associated brain parenchymal abnormalities and important surgical access criteria like normal variants and altered middle ear anatomy. Abnormal imaging findings were seen in 15 of 32 patients, on the basis of which 5 were rejected for implantation; suboptimal results were predicted in 5 other abnormal cases. Side selection and predicting ease of electrode insertion could also be ascertained in unilateral pathologies. Seventeen cases had normal imaging, twelve of which were operated. Eleven of these had uneventful surgery. All three cases with abnormal imaging, taken for surgery under varied risk of suboptimal results had mild and manageable intraoperative complications. Tailored protocol of MRI with BFFE, T2/FLAIR axials and HRCT temporal bone is an indispensable tool and radiologist plays key role in candidate selection and exclusion, surgical planning and warning surgeon about potential intraoperative complications in cochlear implant surgery.

Analysis of the paraspinal muscle morphology of the lumbar spine using a convolutional neural network (CNN)

PurposeThis single-center study aimed to develop a convolutional neural network to segment multiple consecutive axial magnetic resonance imaging (MRI) slices of the lumbar spinal muscles of patients with lower back pain and automatically classify fatty muscle degeneration.MethodsWe developed a fully connected deep convolutional neural network (CNN) with a pre-trained U-Net model trained on a dataset of 3,650 axial T2-weighted MRI images from 100 patients with lower back pain. We included all qualities of MRI; the exclusion criteria were fractures, tumors, infection, or spine implants. The training was performed using k-fold cross-validation (k = 10), and performance was evaluated using the dice similarity coefficient (DSC) and cross-sectional area error (CSA error). For clinical correlation, we used a simplified Goutallier classification (SGC) system with three classes.ResultsThe mean DSC was high for overall muscle (0.91) and muscle tissue segmentation (0.83) but showed deficiencies in fatty tissue segmentation (0.51). The CSA error was small for the overall muscle area of 8.42%, and fatty tissue segmentation showed a high mean CSA error of 40.74%. The SGC classification was correctly predicted in 75% of the patients.ConclusionOur fully connected CNN segmented overall muscle and muscle tissue with high precision and recall, as well as good DSC values. The mean predicted SGC values of all available patient axial slices showed promising results. With an overall Error of 25%, further development is needed for clinical implementation. Larger datasets and training of other model architectures are required to segment fatty tissue more accurately.

Validation of Phase-Resolved Functional Lung (PREFUL) Magnetic Resonance Imaging Pulse Wave Transit Time Compared to Echocardiography in Chronic Obstructive Pulmonary Disease.

Phase-resolved functional lung (PREFUL) magnetic resonance imaging (MRI) pulmonary pulse wave transit time (pPTT) is a contrast agent free, vascular imaging biomarker, but has not been validated in chronic obstructive pulmonary disease (COPD). To validate PREFUL with echocardiographic pPTT as a reference standard and to compare arterial/venous pPTT mapping with spirometry and clinical parameters. Prospective. Twenty-one patients (62% female) with COPD and 44 healthy participants (50% female). 1.5 T; 2D-spoiled gradient-echo sequence. Three coronal PREFUL MRI slices, echocardiography, and spirometry including forced expiratory volume in 1 second (FEV1, liter) and predicted defined as FEV1 in% divided by the population average FEV1%, were performed. Pulmonary pulse transit time from the main artery to the microvasculature (PREFUL pPTT), to the right upper lobe vein (PREFUL pPTTav , echo pPTTav ), from microvasculature to right upper lobe vein (PREFULvein ) and the ratio of PREFUL pPTT to PREFUL pPTTvein were calculated. Body mass index (BMI), Global Initiative for COPD (GOLD) stage 1-4, disease duration, and cigarette packs smoked per day multiplied by the smoked years (pack years) were computed. Shapiro-Wilk-test, paired-two-sided-t-tests, Bland-Altman-analysis, coefficient of variation, Pearson ρ were applied, pPTT data were compared between 21 subjects from the 44 healthy subjects who were age- and sex-matched to the COPD cohort, P < 0.05 was considered statistically significant. PREFUL pPTTav significantly correlated with echo pPTTav (ρ=0.95) with 1.85 msec bias, 95% limits of agreement: 55.94 msec, -52.23 msec in all participants (P=0.59). In the healthy participants, PREFUL and echo pPTTav significantly correlated with age (ρ=0.81, ρ=0.78), FEV1 (ρ=-0.47, ρ=-0.34) and BMI (ρ=0.56, ρ=0.51). In COPD patients, PREFUL pPTT significantly correlated with FEV1 predicted (ρ=-0.59), GOLD (ρ=0.53), disease duration (ρ=0.54), and pack years (ρ=0.49). Arteriovenous PTT measured by PREFUL MRI corresponds precisely to echocardiography and appears to be feasible even in severe COPD. 1 TECHNICAL EFFICACY: Stage 2.

Discovering genetic biomarkers for Alzheimer’s disease using 2D‐CNN and GWAS

AbstractBackgroundTo identify candidate neuroimaging and genetic biomarkers for Alzheimer’s Disease (AD) and other brain diseases, especially for little‐known brain disorders, in a data‐driven way, we advocate an approach which incorporates an adaptive classifier ensemble model by combining Convolutional Neural Network (CNN) and Ensemble Learning with Genetic Algorithm (GA), i.e., the CNN‐EL‐GA method, into Genome‐Wide Association Studies (GWAS).MethodIn the CNN‐EL‐GA method, a large number of base classifiers, i.e., CNN models, were trained utilizing a set of sagittal, coronal, or transversal magnetic resonance imaging slices, and the CNN models with strong discriminability were then selected and integrated into a single classifier ensemble with the GA for classifying AD. While the generalization capability of the acquired classifier ensemble was maximized, the intersection points were decided by the most discriminative slices among the sagittal, coronal, and transversal slice sets. The gray matter volumes of the top ten discriminative brain regions which contained the most intersection points were utilized to carry out GWAS together with the genotypic data. The pipeline of the proposed approach incorporating the CNN‐EL‐GA method into GWAS for discovering candidate genetic biomarkers of AD is shown in Figure 1.ResultSix genes of PCDH11X/Y, TPTE2, LOC107985902, MUC16 and LINC01621 as well as Single Nucleotide Polymorphisms, e.g., rs36088804, rs34640393, rs2451078, rs10496214, rs17016520, rs2591597, rs9352767 and rs5941380, were identified.ConclusionThis approach overcomes the limitations associated with subjective factors while adaptively achieving more robust and effective candidate biomarkers in a data‐driven way. The approach is promising to advance the discovery of effective candidate genetic biomarkers for brain disorders.

Shape registration with learned deformations for 3D shape reconstruction from sparse and incomplete point clouds.

Shape reconstruction from sparse point clouds/images is a challenging and relevant task required for a variety of applications in computer vision and medical image analysis (e.g. surgical navigation, cardiac motion analysis, augmented/virtual reality systems). A subset of such methods, viz. 3D shape reconstruction from 2D contours, is especially relevant for computer-aided diagnosis and intervention applications involving meshes derived from multiple 2D image slices, views or projections. We propose a deep learning architecture, coined Mesh Reconstruction Network (MR-Net), which tackles this problem. MR-Net enables accurate 3D mesh reconstruction in real-time despite missing data and with sparse annotations. Using 3D cardiac shape reconstruction from 2D contours defined on short-axis cardiac magnetic resonance image slices as an exemplar, we demonstrate that our approach consistently outperforms state-of-the-art techniques for shape reconstruction from unstructured point clouds. Our approach can reconstruct 3D cardiac meshes to within 2.5-mm point-to-point error, concerning the ground-truth data (the original image spatial resolution is ∼1.8×1.8×10mm3). We further evaluate the robustness of the proposed approach to incomplete data, and contours estimated using an automatic segmentation algorithm. MR-Net is generic and could reconstruct shapes of other organs, making it compelling as a tool for various applications in medical image analysis.

Lumbar Epidural Lipomatosis is Increased in Patients With Morbid Obesity and Subsequently Decreases After Bariatric Surgery

Spinal epidural lipomatosis (SEL) is characterized by symptomatic neurogenic compression from adipose tissue in the spinal canal. The question arises whether patients with morbid obesity have higher volumes of epidural adipose tissue (EAT) in their lumbar spinal canal compared with patients with a normal weight, and to what extent this decreases after bariatric surgery. In this explorative study the lumbar EAT volume was assessed in 25 patients with morbid obesity (body mass index [BMI] >40) using available lumbar magnetic resonance imaging (MRI) prior to their bariatric surgery. An age- and sex-matched control group (n= 25) of patients with a normal weight (BMI 20-25) was used for comparison. Participants from the obesity group underwent a postoperative control MRI. The mean volume of EAT per MRI slice of the group of patients with obesity was significantly higher than for the group of patients with normal weight (mean: 83.3 ± 30.7 mm3 vs. mean 56.5 ± 25 mm3; P < 0.001). Fifteen participants with obesity (15 of 25) agreed to undergo a follow-up MRI. There was a significant decrease in EAT volume per MRI slice for these 15 participants (mean 82 ± 25.5 mm3 vs. 46 ± 20.0 mm3; P < 0.001) over time up to 3.6 (range: 1.2-6.0) years after bariatric surgery. Patients with obesity have significantly larger volumes of EAT in comparison with patients with normal weight. After bariatric surgery, a significant weight loss coincided with a significant volumetric reduction of this adipose tissue in the spinal canal. Future prospective studies in patients with symptomatic SEL may elucidate whether decreases in EAT volume influence concurrent neurogenic claudication symptoms.

Deep Learning Framework to Detect Ischemic Stroke Lesion in Brain MRI Slices of Flair/DW/T1 Modalities

Ischemic stroke lesion (ISL) is a brain abnormality. Studies proved that early detection and treatment could reduce the disease impact. This research aimed to develop a deep learning (DL) framework to detect the ISL in multi-modality magnetic resonance image (MRI) slices. It proposed a convolutional neural network (CNN)-supported segmentation and classification to execute a consistent disease detection framework. The developed framework consisted of the following phases; (i) visual geometry group (VGG) developed VGG16 scheme supported SegNet (VGG-SegNet)-based ISL mining, (ii) handcrafted feature extraction, (iii) deep feature extraction using the chosen DL scheme, (iv) feature ranking and serial feature concatenation, and (v) classification using binary classifiers. Fivefold cross-validation was employed in this work, and the best feature was selected as the final result. The attained results were separately examined for (i) segmentation; (ii) deep-feature-based classification, and (iii) concatenated feature-based classification. The experimental investigation is presented using the Ischemic Stroke Lesion Segmentation (ISLES2015) database. The attained result confirms that the proposed ISL detection framework gives better segmentation and classification results. The VGG16 scheme helped to obtain a better result with deep features (accuracy &gt; 97%) and concatenated features (accuracy &gt; 98%).

A canonical visualization tool for SEEG electrodes.

Stereoencephalographic (SEEG) electrodes are clinically implanted into the brains of patients with refractory epilepsy to locate foci of seizure onset. They are increasingly used in neurophysiology research to determine focal human brain activity in response to tasks or stimuli. Clear visualization of SEEG electrode location with respect to patient anatomy on magnetic resonance image (MRI) scan is vital to neuroscientific understanding. An intuitive way to accomplish this is to plot brain activity and labels at electrode locations on closest MRI slices along the canonical axial, coronal, and sagittal planes. Therefore, we've developed an open-source software tool in Matlab for visualizing SEEG electrode positions, determined from computed tomography (CT), onto canonical planes of resliced brain MRI. The code and graphical user interface are available at: https://github.com/MultimodalNeuroimagingLab/mnl_seegviewClinical Relevance- This tool enables precise communication of SEEG electrode activity and location by visualization on slices of MRI in canonical axial, coronal, and sagittal planes.

Optimal Method for Fetal Brain Age Prediction Using Multiplanar Slices From Structural Magnetic Resonance Imaging.

The accurate prediction of fetal brain age using magnetic resonance imaging (MRI) may contribute to the identification of brain abnormalities and the risk of adverse developmental outcomes. This study aimed to propose a method for predicting fetal brain age using MRIs from 220 healthy fetuses between 15.9 and 38.7 weeks of gestational age (GA). We built a 2D single-channel convolutional neural network (CNN) with multiplanar MRI slices in different orthogonal planes without correction for interslice motion. In each fetus, multiple age predictions from different slices were generated, and the brain age was obtained using the mode that determined the most frequent value among the multiple predictions from the 2D single-channel CNN. We obtained a mean absolute error (MAE) of 0.125 weeks (0.875 days) between the GA and brain age across the fetuses. The use of multiplanar slices achieved significantly lower prediction error and its variance than the use of a single slice and a single MRI stack. Our 2D single-channel CNN with multiplanar slices yielded a significantly lower stack-wise MAE (0.304 weeks) than the 2D multi-channel (MAE = 0.979, p < 0.001) and 3D (MAE = 1.114, p < 0.001) CNNs. The saliency maps from our method indicated that the anatomical information describing the cortex and ventricles was the primary contributor to brain age prediction. With the application of the proposed method to external MRIs from 21 healthy fetuses, we obtained an MAE of 0.508 weeks. Based on the external MRIs, we found that the stack-wise MAE of the 2D single-channel CNN (0.743 weeks) was significantly lower than those of the 2D multi-channel (1.466 weeks, p < 0.001) and 3D (1.241 weeks, p < 0.001) CNNs. These results demonstrate that our method with multiplanar slices accurately predicts fetal brain age without the need for increased dimensionality or complex MRI preprocessing steps.

Super-resolution Reconstruction MRI Application in Fetal Neck Masses and Congenital High Airway Obstruction Syndrome.

ObjectiveReliable airway patency diagnosis in fetal tracheolaryngeal obstruction is crucial to select and plan ex utero intrapartum treatment (EXIT) surgery. We compared the clinical utility of magnetic resonance imaging (MRI) super-resolution reconstruction (SRR) of the trachea, which can mitigate unpredictable fetal motion effects, with standard 2-dimensional (2D) MRI for airway patency diagnosis and assessment of fetal neck mass anatomy.Study DesignA single-center case series of 7 consecutive singleton pregnancies with complex upper airway obstruction (2013-2019).SettingA tertiary fetal medicine unit performing EXIT surgery.MethodsMRI SRR of the trachea was performed involving rigid motion correction of acquired 2D MRI slices combined with robust outlier detection to reconstruct an isotropic high-resolution volume. SRR, 2D MRI, and paired data were blindly assessed by 3 radiologists in 3 experimental rounds.ResultsAirway patency was correctly diagnosed in 4 of 7 cases (57%) with 2D MRI as compared with 2 of 7 cases (29%) with SRR alone or paired 2D MRI and SRR. Radiologists were more confident (P = .026) in airway patency diagnosis when using 2D MRI than SRR. Anatomic clarity was higher with SRR (P = .027) or paired data (P = .041) in comparison with 2D MRI alone. Radiologists detected further anatomic details by using paired images versus 2D MRI alone (P < .001). Cognitive load, as assessed by the NASA Task Load Index, was increased with paired or SRR data in comparison with 2D MRI.ConclusionThe addition of SRR to 2D MRI does not increase fetal airway patency diagnostic accuracy but does provide improved anatomic information, which may benefit surgical planning of EXIT procedures.

3D shearlet-based descriptors combined with deep features for the classification of Alzheimer's disease based on MRI data

Alzheimer's disease (AD) is a neurodegenerative disease that afflicts millions of people worldwide. Early detection of AD is critical, as drug trials show a promising advantage to those patients with early diagnoses. In this study, magnetic resonance imaging (MRI) datasets from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and The Open Access Series of Imaging Studies are used. Our method for performing the classification of AD is to combine a set of shearlet-based descriptors with deep features. A major challenge in classifying such MRI datasets is the high dimensionality of feature vectors because of the large number of slices of each MRI sample. Given the volumetric nature of the MRI data, we propose using the 3D shearlet transform (3D-ST), but we obtain the average of all directionalities, which reduces the dimensionality. On the other hand, we propose to leverage the capabilities of convolutional neural networks (CNN) to learn feature maps from stacked MRI slices, which generate a very compact feature vector for each MRI sample. The 3D-ST and CNN feature vectors are combined for the classification of AD. After the concatenation of the feature vectors, they are used to train a classifier. Alternatively, a custom CNN model is utilized, in which the descriptors are further processed end to end to obtain the classification model. Our experimental results show that the fusion of shearlet-based descriptors and deep features improves classification performance, especially on the ADNI dataset.

Quality Control after Intracochlear Intralabyrinthine Schwannoma Resection and Cochlear Implantation.

Background: The combination of intralabyrinthine schwannoma (ILS) removal and cochlear implantation is the standard of care after surgical resection for audiological rehabilitation. Intracochlear ILS is not only the most frequent tumor in this group of schwannomas, but it is also, to some degree, surgically the most challenging because of its position behind the modiolus. Recent developments in the knowledge of implant position, implant magnet choice, and magnetic resonance imaging (MRI) sequences make an MRI follow-up after surgical removal possible. Thus far, no results are known about the surgical success and residual rate of these kind of tumors. The aim of the present study was to perform an early MRI follow-up for the evaluation of residual or recurrent intracochlear ILS after surgical removal and cochlear implantation. Methods: In a retrospective study, we evaluated seven patients after an intracochlear ILS removal and single-stage cochlear implantation with a mean period of 13.4 months post surgery with a 3T T1 GAD 2 mm sequence for a residual ILS. Patients were operated on using an individualized technique concept. Results: In six out of seven cases, 3 T T1 GAD 2 mm MRI follow-up showed no residual or recurrent tumor. In one case, a T1 signal indicated a tumor of the upper inner auditory canal (IAC) at the MRI follow up. Conclusion: MRI follow-up as a quality control tool after ILS removal and cochlear implantation is highly important to exclude residual tumors. Long-term MRI evaluation results are needed and can be obtained under consideration of implant position, implant magnet, and MRI sequence choice. A preoperative MRI slice thickness less than 2 mm can be recommended to visualize possible modiolar and IAC expansion.

MRI-based three-dimensional reconstruction for staging cervical cancer and predicting high-risk patients.

BackgroundCervical tumors usually have an irregular morphology. It is often difficult to estimate tumor size or volume based on a diameter measurement from a two-dimensional magnetic resonance imaging slice. This study aimed to explore the use of magnetic resonance imaging-based three-dimensional reconstruction in cervical cancer.MethodsWe retrospectively created a three-dimensional reconstruction based on the pre-treatment magnetic resonance imaging data of 54 cervical cancer patients at a single center to evaluate tumor size and extent of invasion, as well as to review cervical cancer staging and treatment. The tissues and organs were automatically outlined by the three-dimensional application, based on the signal intensity difference of magnetic resonance imaging data.ResultsThe maximum tumor diameters calculated using the magnetic resonance imaging-based three-dimensional reconstruction were larger than those calculated from the direct magnetic resonance imaging findings or gynecological examinations. Initial underestimation of the maximum tumor diameter led to under-staging in up to 29.6% of patients. The magnetic resonance imaging-based three-dimensional reconstruction revealed that upstaging was warranted based on lymph node metastasis (3.7% of patients) and invasion of the vaginal fornix (1.9% of patients). Lymph node metastasis was associated with a significantly larger tumor volume (P<0.05). A volume cut-off value ≥18.6 mL provided 60% sensitivity, 96.7% specificity, 75% positive predictive value and 93.5% negative predictive value for predicting high-risk patients (P<0.05).ConclusionsMagnetic resonance imaging-based three-dimensional reconstruction is a new approach that could potentially measure cervical cancer more accurately.

Brain Tumor Detection and Classification on MR Images by a Deep Wavelet Auto-Encoder Model.

The process of diagnosing brain tumors is very complicated for many reasons, including the brain’s synaptic structure, size, and shape. Machine learning techniques are employed to help doctors to detect brain tumor and support their decisions. In recent years, deep learning techniques have made a great achievement in medical image analysis. This paper proposed a deep wavelet autoencoder model named “DWAE model”, employed to divide input data slice as a tumor (abnormal) or no tumor (normal). This article used a high pass filter to show the heterogeneity of the MRI images and their integration with the input images. A high median filter was utilized to merge slices. We improved the output slices’ quality through highlight edges and smoothened input MR brain images. Then, we applied the seed growing method based on 4-connected since the thresholding cluster equal pixels with input MR data. The segmented MR image slices provide two two-layer using the proposed deep wavelet auto-encoder model. We then used 200 hidden units in the first layer and 400 hidden units in the second layer. The softmax layer testing and training are performed for the identification of the MR image normal and abnormal. The contribution of the deep wavelet auto-encoder model is in the analysis of pixel pattern of MR brain image and the ability to detect and classify the tumor with high accuracy, short time, and low loss validation. To train and test the overall performance of the proposed model, we utilized 2500 MR brain images from BRATS2012, BRATS2013, BRATS2014, BRATS2015, 2015 challenge, and ISLES, which consists of normal and abnormal images. The experiments results show that the proposed model achieved an accuracy of 99.3%, loss validation of 0.1, low FPR and FNR values. This result demonstrates that the proposed DWAE model can facilitate the automatic detection of brain tumors.

Discovery of Genetic Biomarkers for Alzheimer's Disease Using Adaptive Convolutional Neural Networks Ensemble and Genome-Wide Association Studies.

To identify candidate neuroimaging and genetic biomarkers for Alzheimer's disease (AD) and other brain disorders, especially for little-investigated brain diseases, we advocate a data-driven approach which incorporates an adaptive classifier ensemble model acquired by integrating Convolutional Neural Network (CNN) and Ensemble Learning (EL) with Genetic Algorithm (GA), i.e., the CNN-EL-GA method, into Genome-Wide Association Studies (GWAS). Above all, a large number of CNN models as base classifiers were trained using coronal, sagittal, or transverse magnetic resonance imaging slices, respectively, and the CNN models with strong discriminability were then selected to build a single classifier ensemble with the GA for classifying AD, with the help of the CNN-EL-GA method. While the acquired classifier ensemble exhibited the highest generalization capability, the points of intersection were determined with the most discriminative coronal, sagittal, and transverse slices. Finally, we conducted GWAS on the genotype data and the phenotypes, i.e., the gray matter volumes of the top ten most discriminative brain regions, which contained the ten most points of intersection. Six genes of PCDH11X/Y, TPTE2, LOC107985902, MUC16 and LINC01621 as well as Single-Nucleotide Polymorphisms, e.g., rs36088804, rs34640393, rs2451078, rs10496214, rs17016520, rs2591597, rs9352767 and rs5941380, were identified. This approach overcomes the limitations associated with the impact of subjective factors and dependence on prior knowledge while adaptively achieving more robust and effective candidate biomarkers in a data-driven way. The approach is promising to facilitate discovering effective candidate genetic biomarkers for brain disorders, as well as to help improve the effectiveness of identified candidate neuroimaging biomarkers for brain diseases.

BRAIN STROKE DETECTION USING TENSOR FACTORIZATION AND MACHINE LEARNING MODELS

Stroke is one of the foremost common disorders among the elderly. Early detection of stroke from Magnetic Resonance Imaging (MRI) is typically based on the representation method of these images. Representing MRI slices in two dimensional structures (matrices) implies ignoring the dependencies between these slices. Additionally, to combine all features exist in these slices requires more computations and time. However, this results in inexact diagnosis. In this paper, we propose a new tensor-based approach for stroke detection from MRI. The proposed methodology has two phases. In first phase, each patient’s MRI are represented as a tensor. Tensor representations are powerful because they capture the dependencies in high-dimensional data, MRI of patient, which gives more reliable and accurate results. Also, tensor factorization is used as a method for feature extraction and reduction, which improves the performance and accuracy of classifiers. In second phase, these extracted features are used to train support vector machine (SVM) and XGBoost classifiers to classify MRI images into normal and abnormal. The proposed method is assessed with MRI dataset, and the conducted experiments illustrate the efficiency of this approach. It achieves classification accuracy of 98%.

Convolutional Neural Networks for Automated Classification of Prostate Multiparametric Magnetic Resonance Imaging Based on Image Quality.

BackgroundProstate magnetic resonance imaging (MRI) is technically demanding, requiring high image quality to reach its full diagnostic potential. An automated method to identify diagnostically inadequate images could help optimize image quality.PurposeTo develop a convolutional neural networks (CNNs) based analysis pipeline for the classification of prostate MRI image quality.Study TypeRetrospective.SubjectsThree hundred sixteen prostate mpMRI scans and 312 men (median age 67).Field Strength/SequenceA 3 T; fast spin echo T2WI, echo planar imaging DWI, ADC, gradient‐echo dynamic contrast enhanced (DCE).AssessmentMRI scans were reviewed by three genitourinary radiologists (V.P., M.D.M., S.C.) with 21, 12, and 5 years of experience, respectively. Sequences were labeled as high quality (Q1) or low quality (Q0) and used as the reference standard for all analyses.Statistical TestsSequences were split into training, validation, and testing sets (869, 250, and 120 sequences, respectively). Inter‐reader agreement was assessed with the Fleiss kappa. Following preprocessing and data augmentation, 28 CNNs were trained on MRI slices for each sequence. Model performance was assessed on both a per‐slice and a per‐sequence basis. A pairwise t‐test was performed to compare performances of the classifiers.ResultsThe number of sequences labeled as Q0 or Q1 was 38 vs. 278 for T2WI, 43 vs. 273 for DWI, 41 vs. 275 for ADC, and 38 vs. 253 for DCE. Inter‐reader agreement was almost perfect for T2WI and DCE and substantial for DWI and ADC. On the per‐slice analysis, accuracy was 89.95% ± 0.02% for T2WI, 79.83% ± 0.04% for DWI, 76.64% ± 0.04% for ADC, 96.62% ± 0.01% for DCE. On the per‐sequence analysis, accuracy was 100% ± 0.00% for T2WI, DWI, and DCE, and 92.31% ± 0.00% for ADC. The three best algorithms performed significantly better than the remaining ones on every sequence (P‐value < 0.05).Data ConclusionCNNs achieved high accuracy in classifying prostate MRI image quality on an individual‐slice basis and almost perfect accuracy when classifying the entire sequences.Evidence Level4Technical EfficacyStage 1

Extraction and Evaluation of Corpus Callosum from 2D Brain MRI Slice: A Study with Cuckoo Search Algorithm.

The work proposes a computer-based diagnosis method (CBDM) to delineate and assess the corpus callosum (CC) segment from the 2-dimensional (2D) brain magnetic resonance images (MRI). The proposed CBDM consists of two parts: (1) preprocessing and (2) postprocessing sections. The preprocessing tools have a multithreshold technique with the chaotic cuckoo search (CCS) algorithm and a preferred threshold procedure. The postprocessing employs a delineation process for extracting the CC section. The proposed CBDM finally extracts the vital CC parameters, such as total brain area (TBA) and CC area (CCA) to classify the considered 2D MRI slices into the control and autism spectrum disorder (ASD) groups. This attempt considers the benchmark brain MRI database which includes ABIDE and MIDAS for the experimental investigation. The results obtained with ABIDE dataset are further confirmed against the fuzzy C-means driven level set (FCM + LS) and multiphase level set (MLS) technique and the proposed CBDM with Shannon entropy along with active contour (SE + AC) presented improved result in comparison to the existing methodologies. Further, the performance of CBDM is confirmed on MIDAS and clinical dataset. The experimental outcomes approve that the proposed CBDM extracts the CC section from the 2D MR brain images that have higher accuracy compared to alternative techniques.

The Effect of Aberrant Rotation on Radiographic Patellar Height Measurement Using Canton-Deschamps Index: A Cadaveric Analysis.

The Caton-Deschamps Index (CDI) is a measurement used to evaluate patella alta based on true lateral radiographs; however, no prior study has investigated how altering the degree of radiograph aberrancy affects CDI measurement. The primary and secondary purpose of this study was to evaluate effects of rotational radiographic changes on patella height measurements and compare these findings to MRI measurements, respectively. Five cadaver knees (n = 5) were utilized in this study. True lateral radiographs were obtained for each specimen by using a fluoroscopic C-arm machine. The C-arm was then altered in two planes (axial and coronal) in both the clockwise and counterclockwise direction and radiographs were taken at 5, 10, and 15 degrees of error from the true lateral position. A CDI measurement of each specimen was performed based on sagittal magnetic resonance imaging (MRI) slices and compared with radiographic CDI measurements. Three orthopedic surgeons measured the CDI for each radiograph and MRI performed. Interrater reliability and changes in CDI were analyzed. Clinically significant difference in CDI was set to 0.1. Mean intraclass correlation coefficient was high (≥0.7) at true lateral and at all varying degrees of error. When performing a pairwise comparison of mean CDI from the true lateral position to increasing degrees of error, statistically significant differences were observed in the axial plane. The largest change in CDI measurements was seen with rotational malposition in the axial plane and counterclockwise direction. No statistically significant differences in mean CDI were observed in the coronal plane. The change in CDI from the true lateral position reached an absolute maximum of at least 0.1 in all four scenarios at each tested degree of error. This study found that aberrant radiographic rotation in the axial plane resulted in a significantly different mean CDI measurement when compared with true lateral radiographs. All degrees of error in both directions and in both planes could have a clinically significant effect on CDI (≥0.1). Our findings confirm the importance of a perfect true lateral radiograph when measuring patella height.