An Improved Brain MRI Classification Methodology Based on Statistical Features and Machine Learning Algorithms.

In this paper, we have proposed a novel methodology based on statistical features and different machine learning algorithms. The proposed model can be divided into three main stages, namely, preprocessing, feature extraction, and classification. In the preprocessing stage, the median filter has been used in order to remove salt-and-pepper noise because MRI images are normally affected by this type of noise, the grayscale images are also converted to RGB images in this stage. In the preprocessing stage, the histogram equalization has also been used to enhance the quality of each RGB channel. In the feature extraction stage, the three channels, namely, red, green, and blue, are extracted from the RGB images and statistical measures, namely, mean, variance, skewness, kurtosis, entropy, energy, contrast, homogeneity, and correlation, are calculated for each channel; hence, a total of 27 features, 9 for each channel, are extracted from an RGB image. After the feature extraction stage, different machine learning algorithms, such as artificial neural network, k-nearest neighbors' algorithm, decision tree, and Naïve Bayes classifiers, have been applied in the classification stage on the features extracted in the feature extraction stage. We recorded the results with all these algorithms and found that the decision tree results are better as compared to the other classification algorithms which are applied on these features. Hence, we have considered decision tree for further processing. We have also compared the results of the proposed method with some well-known algorithms in terms of simplicity and accuracy; it was noted that the proposed method outshines the existing methods.

Preoperative MRI Classification May Not Predict Symptom Relief after Uterine Artery Embolization in Patients with Adenomyosis.

OBJECTIVE: To investigate the association between magnetic resonance imaging (MRI) classification and symptom relief after uterine artery embolization (UAE) in patients with adenomyosis. METHODS: Totally, 73 patients with symptomatic adenomyosis who underwent UAE were retrospectively analyzed. Preoperative MRI classification was defined as: type I, high signal on both T2-weighted images (T2WI) and T1-weighted images (T1WI); type III, high signal only on T2WI, and type II, high signal on neither T1WI nor T2WI. Dysmenorrhea was measured with the visual-analog scales and the degree of menorrhagia was measured according to the number of sanitary pads used in one menstrual cycle. Dysmenorrhea and menorrhagia were measured before UAE and 12 months after UAE. RESULTS: The number of the type I, II, III cases was 23, 37, and 13, respectively. The baseline characteristics of the three groups exhibited no significant difference. The alleviation rates of dysmenorrhea among type I, II, III cases were 73.9%, 89.2%, and 84.6%, respectively (P=0.455). The alleviation rates of menorrhagia for type I, II, III were 69.6%, 78.4%, and 92.3%, respectively (P=0.714). CONCLUSION: Pre-procedure MRI classification and symptom relief after UAE exhibited no significant association. UAE has a favorable mid-term control on dysmenorrhea and menorrhagia among patients with adenomyosis. Preoperative MRI classification might not indicate symptom relief. More research is needed before changing clinical practice.

Random walks on B distributed resting-state functional connectivity to identify Alzheimer's disease and Mild Cognitive Impairment.

OBJECTIVE: Resting-state functional connectivity reveals a promising way for the early detection of dementia. This study proposes a novel method to accurately classify Healthy Controls, Early Mild Cognitive Impairment, Late Mild Cognitive Impairment, and Alzheimer's Disease individuals. METHODS: A novel mapping function based on the B distribution has been developed to map correlation matrices to robust functional connectivity. The node2vec algorithm is applied to the functional connectivity to produce node embeddings. The concatenation of these embedding has been used to derive the patients' feature vectors for further feeding into the Support Vector Machine and Logistic Regression classifiers. RESULTS: The experimental results indicate promising results in the complex four-class classification problem with an accuracy rate of 97.73% and a quadratic kappa score of 96.86% for the Support Vector Machine. These values are 97.32% and 96.74% for Logistic Regression. CONCLUSION: This study presents an accurate automated method for dementia classification. Default Mode Network and Dorsal Attention Network have been found to demonstrate a significant role in the classification method. SIGNIFICANCE: A new mapping function is proposed in this study, the mapping function improves accuracy by 10-11% in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.

Brain tissues have single-voxel signatures in multi-spectral MRI.

Since the seminal works by Brodmann and contemporaries, it is well-known that different brain regions exhibit unique cytoarchitectonic and myeloarchitectonic features. Transferring the approach of classifying brain tissues - and other tissues - based on their intrinsic features to the realm of magnetic resonance (MR) is a longstanding endeavor. In the 1990s, atlas-based segmentation replaced earlier multi-spectral classification approaches because of the large overlap between the class distributions. Here, we explored the feasibility of performing global brain classification based on intrinsic MR features, and used several technological advances: ultra-high field MRI, q-space trajectory diffusion imaging revealing voxel-intrinsic diffusion properties, chemical exchange saturation transfer and semi-solid magnetization transfer imaging as a marker of myelination and neurochemistry, and current neural network architectures to analyze the data. In particular, we used the raw image data as well to increase the number of input features. We found that a global brain classification of roughly 97 brain regions was feasible with gross classification accuracy of 60%; and that mapping from voxel-intrinsic MR data to the brain region to which the data belongs is possible. This indicates the presence of unique MR signals of different brain regions, similar to their cytoarchitectonic and myeloarchitectonic fingerprints.

A novel pipeline leveraging surface-based features of small subcortical structures to classify individuals with autism spectrum disorder.

Autism spectrum disorder (ASD) is accompanied with widespread impairment in social-emotional functioning. Classification of ASD using sensitive morphological features derived from structural magnetic resonance imaging (MRI) of the brain may help us to better understand ASD-related mechanisms and improve related automatic diagnosis. Previous studies using T1 MRI scans in large heterogeneous ABIDE dataset with typical development (TD) controls reported poor classification accuracies (around 60%). This may because they only considered surface-based morphometry (SBM) as scalar estimates (such as cortical thickness and surface area) and ignored the neighboring intrinsic geometry information among features. In recent years, the shape-related SBM achieves great success in discovering the disease burden and progression of other brain diseases. However, when focusing on local geometry information, its high dimensionality requires careful treatment in its application to machine learning. To address the above challenges, we propose a novel pipeline for ASD classification, which mainly includes the generation of surface-based features, patch-based surface sparse coding and dictionary learning, Max-pooling and ensemble classifiers based on adaptive optimizers. The proposed pipeline may leverage the sensitivity of brain surface morphometry statistics and the efficiency of sparse coding and Max-pooling. By introducing only the surface features of bilateral hippocampus that derived from 364 male subjects with ASD and 381 age-matched TD males, this pipeline outperformed five recent MRI-based ASD classification studies with >80% accuracy in discriminating individuals with ASD from TD controls. Our results suggest shape-related SBM features may further boost the classification performance of MRI between ASD and TD.

Relationship between Machine-Learning Image Classification of T2-Weighted Intramedullary Hypointensity on 3 Tesla Magnetic Resonance Imaging and Clinical Outcome in Dogs with Severe Spinal Cord Injury.

Early prognostic information in cases of severe spinal cord injury can aid treatment planning and stratification for clinical trials. Analysis of intraparenchymal signal change on magnetic resonance imaging has been suggested to inform outcome prediction in traumatic spinal cord injury. We hypothesized that intraparenchymal T2-weighted hypointensity would be associated with a lower potential for functional recovery and a higher risk of progressive neurological deterioration in dogs with acute, severe, naturally occurring spinal cord injury. Our objectives were to: 1) demonstrate capacity for machine-learning criteria to identify clinically relevant regions of hypointensity and 2) compare clinical outcomes for cases with and without such regions. A total of 95 dogs with complete spinal cord injury were evaluated. An image classification system, based on Speeded-Up Robust Features (SURF), was trained to recognize individual axial T2-weighted slices that contained hypointensity. The presence of such slices in a given transverse series was correlated with a lower chance of functional recovery (odds ratio [OR], 0.08; confidence interval [CI], 0.02-0.38; p < 10-3) and with a higher risk of neurological deterioration (OR, 0.14; 95% CI, 0.05-0.42; p < 10-3). Identification of intraparenchymal T2-weighted hypointensity in severe, naturally occurring spinal cord injury may be assisted by an image classification tool and is correlated with functional recovery.

Subclassification of BI-RADS 4 Magnetic Resonance Lesions: A Systematic Review and Meta-Analysis.

OBJECTIVE: This research aims to investigate and evaluate the diagnostic efficacy of magnetic resonance imaging (MRI) in classifying Breast Imaging Reporting and Data System (BI-RADS) 4 lesions into subcategories: 4a, 4b, and 4c, so as to limit biopsies of suspected lesions in the breast. METHODS: The PubMed, Web of Science, Embase, and Cochrane Library foreign language databases were searched for literature published between January 2000 and July 2018. After analyzing the selection, data extraction, and quality assessment, a meta-analysis was performed, including data pooling, heterogeneity testing, and meta-regression. RESULTS: Fourteen articles, including 18 studies, met the inclusion criteria. The diagnostic efficacy of MRI for BI-RADS 4-weighted summary assay sensitivity and specificity were estimated at 0.95 [95% confidence interval (CI), 0.89-0.98] and 0.87 (95% CI, 0.81-0.91), respectively. The positive and negative likelihood ratios were 7.1 (95% CI, 4.7-10.7) and 0.06 (95% CI, 0.02-0.14), respectively. The diagnostic odds ratio was 126 (95% CI, 37-426), and the area under the receiver operating characteristic curve was 0.95 (95% CI, 0.93-0.97). The malignancy ratio of BI-RADS 4a, 4b, and 4c and malignancy range were 2.5% to 18.3%, 23.5% to 57.1%, and 58.0% to 95.2%, respectively. CONCLUSION: Risk stratification of suspected lesions (BI-RADS categories 4a, 4b, and 4c) can be achieved by MRI. The MRI is an effective auxiliary tool to further subclassify BI-RADS 4 lesions and prevent unnecessary biopsy of BI-RADS 4a lesions.

Gadoxetic acid-enhanced MRI of hepatocellular carcinoma: Diagnostic performance of category-adjusted LR-5 using modified criteria.

The Liver Imaging Reporting and Data System (LI-RADS) is widely adopted for the noninvasive diagnosis of hepatocellular carcinoma (HCC). Herein, possible strategies to improve the diagnostic performance of LR-5 without reducing specificity for HCC were investigated. This retrospective study included 792 patients who underwent gadoxetate disodium-enhanced magnetic resonance imaging. Hepatic observations were categorized according to LI-RADS v2018 and categories were readjusted by upgrading LR4 to LR5 using ancillary features, arterial phase hyperenhancement (APHE) interpreted with subtraction images, indication of no washout when APHE was absent, extension of washout to the transitional phase, and subthreshold growth as a major feature. Based on LI-RADS v2018, LR-5 showed a sensitivity of 71.9% and a specificity of 97.9% for the diagnosis of HCC. Category-readjusted LR-5 after upgrading LR-4 to LR-5 using ancillary features favoring HCC in particular, subthreshold growth as a major feature, extending washout to transitional phase and APHE interpreted using subtraction images showed significantly increased sensitivity (P<0.001) without decreased specificity (Ps>0.05). The sensitivity of LR-5 can be improved without loss of specificity via category readjustment using AFs favoring HCC in particular, subthreshold growth as a major feature, extending washout to transitional phase and APHE interpreted with subtraction images.

Definitions and classification of malformations of cortical development: practical guidelines.

Malformations of cortical development are a group of rare disorders commonly manifesting with developmental delay, cerebral palsy or seizures. The neurological outcome is extremely variable depending on the type, extent and severity of the malformation and the involved genetic pathways of brain development. Neuroimaging plays an essential role in the diagnosis of these malformations, but several issues regarding malformations of cortical development definitions and classification remain unclear. The purpose of this consensus statement is to provide standardized malformations of cortical development terminology and classification for neuroradiological pattern interpretation. A committee of international experts in paediatric neuroradiology prepared systematic literature reviews and formulated neuroimaging recommendations in collaboration with geneticists, paediatric neurologists and pathologists during consensus meetings in the context of the European Network Neuro-MIG initiative on Brain Malformations (https://www.neuro-mig.org/). Malformations of cortical development neuroimaging features and practical recommendations are provided to aid both expert and non-expert radiologists and neurologists who may encounter patients with malformations of cortical development in their practice, with the aim of improving malformations of cortical development diagnosis and imaging interpretation worldwide.

Analysis of magnetic resonance imaging features of ovarian thecoma.

To investigate the magnetic resonance imaging (MRI) findings in ovarian thecoma and improve preoperative diagnostic accuracy.Retrospective analysis was performed on 45 patients with surgically and pathologically confirmed ovarian thecoma. Patients were grouped into those with maximum lesion diameter ≥5 cm and <5 cm. Diagnostic scores (up to 6 points) were evaluated on the basis of MRI performance.The ≥5 cm group contained 36 cases (cystic necrosis, 32 cases) with the following findings: T1WI: isointense signal, 22 cases; slightly hypointense signal, 14 cases; T2WI: isointense signal, 6 cases; slightly hypointense signal, 21 cases; slightly hyperintense signal, 9 cases; Diffusion-weighted imaging (DWI): hyperintense signal, 23 cases; mixed hyperintense signal, 13 cases; slight enhancement on dynamic enhanced scans; pelvic fluid accumulation, 31 cases. The diagnostic score evaluations yielded 6 points in 31 cases, 5 points in 1 case, 4 points in 2 cases, and 3 points in 2 cases. The <5 cm group contained 9 cases (cystic necrosis, 3 cases) with the following findings: T1WI: isointense signal, 3 cases; slightly hypointense signal, 6 cases; T2WI: isointense signal, 2 cases; slightly hypointense signal, 4 cases; slightly hyperintense signal, 3 cases; DWI, hyperintense signal; slight enhancement in 8 cases and significant enhancement in 1 case; pelvic fluid accumulation, 4 cases. The diagnostic score evaluations yielded 6 points in 3 cases, 5 points in 1 case, 4 points in 4 cases, and 3 points in 1 case. (iii) Incidence of pelvic fluid accumulation and cystic necrosis differed depending on the size of the lesion (P = .007, .000).Larger lesions show hyperintense or mixed hyperintense signals on DWI along with pelvic fluid and cystic necrosis; whereas, smaller lesions show a hyperintense signal on DWI, cystic necrosis is rare. MRI characteristics along with the patient age and laboratory findings can improve the accuracy of preoperative diagnosis of these lesions.

Radiogenomic Analysis of Breast Cancer by Linking MRI Phenotypes with Tumor Gene Expression.

Background Better understanding of the molecular biology associated with MRI phenotypes may aid in the diagnosis and treatment of breast cancer. Purpose To discover the associations between MRI phenotypes of breast cancer and their underlying molecular biology derived from gene expression data. Materials and Methods This is a secondary analysis of the Multimodality Analysis and Radiologic Guidance in Breast-Conserving Therapy, or MARGINS, study. MARGINS included patients eligible for breast-conserving therapy between November 2000 and December 2008 for preoperative breast MRI. Tumor RNA was collected for sequencing from surgical specimen. Twenty-one computer-generated MRI features of tumors were condensed into seven MRI factors related to tumor size, shape, initial enhancement, late enhancement, smoothness of enhancement, sharpness, and sharpness variation. These factors were associated with gene expression levels from RNA sequencing by using gene set enrichment analysis. Statistical significance of these associations was evaluated by using a sample permutation test and the false discovery rate. Results Gene expression and MRI data were obtained for 295 patients (mean age, 56 years ± 10.3 [standard deviation]). Larger and more irregular tumors showed increased expression of cell cycle and DNA damage checkpoint genes (false discovery rate <0.25; normalized enrichment statistic [NES], 2.15). Enhancement and sharpness of the tumor margin were associated with expression of ribosomal proteins (false discovery rate <0.25; NES, 1.95). Smoothness of enhancement, tumor size, and tumor shape were associated with expression of genes involved in the extracellular matrix (false discovery rate <0.25; NES, 2.25). Conclusion Breast cancer MRI phenotypes were related to their underlying molecular biology revealed by using RNA sequencing. The association between enhancements and sharpness of the tumor margin with the ribosome suggests that these MRI features may be imaging biomarkers for drugs targeting the ribosome. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Cho in this issue.

Central reader evaluation of MRI scans of the sacroiliac joints from the ASAS classification cohort: discrepancies with local readers and impact on the performance of the ASAS criteria.

OBJECTIVES: The Assessment of SpondyloArthritis international Society (ASAS) MRI working group conducted a multireader exercise on MRI scans from the ASAS classification cohort to assess the spectrum and evolution of lesions in the sacroiliac joint and impact of discrepancies with local readers on numbers of patients classified as axial spondyloarthritis (axSpA). METHODS: Seven readers assessed baseline scans from 278 cases and 8 readers assessed baseline and follow-up scans from 107 cases. Agreement for detection of MRI lesions between central and local readers was assessed descriptively and by the kappa statistic. We calculated the number of patients classified as axSpA by the ASAS criteria after replacing local detection of active lesions by central readers and replacing local reader radiographic sacroiliitis by central reader structural lesions on MRI. RESULTS: Structural lesions, especially erosions, were as frequent as active lesions (≈40%), the majority of patients having both types of lesions. The ASAS definitions for active MRI lesion typical of axSpA and erosion were comparatively discriminatory between axSpA and non-axSpA. Local reader overcall for active MRI lesions was about 30% but this had a minor impact on the number of patients (6.4%) classified as axSpA. Substitution of radiography with MRI structural lesions also had little impact on classification status (1.4%). CONCLUSION: Despite substantial discrepancy between central and local readers in interpretation of both types of MRI lesion, this had a minor impact on the numbers of patients classified as axSpA supporting the robustness of the ASAS criteria for differences in assessment of imaging.

Brain MRI analysis using a deep learning based evolutionary approach.

Convolutional neural network (CNN) models have recently demonstrated impressive performance in medical image analysis. However, there is no clear understanding of why they perform so well, or what they have learned. In this paper, a three-dimensional convolutional neural network (3D-CNN) is employed to classify brain MRI scans into two predefined groups. In addition, a genetic algorithm based brain masking (GABM) method is proposed as a visualization technique that provides new insights into the function of the 3D-CNN. The proposed GABM method consists of two main steps. In the first step, a set of brain MRI scans is used to train the 3D-CNN. In the second step, a genetic algorithm (GA) is applied to discover knowledgeable brain regions in the MRI scans. The knowledgeable regions are those areas of the brain which the 3D-CNN has mostly used to extract important and discriminative features from them. For applying GA on the brain MRI scans, a new chromosome encoding approach is proposed. The proposed framework has been evaluated using ADNI (including 140 subjects for Alzheimer's disease classification) and ABIDE (including 1000 subjects for Autism classification) brain MRI datasets. Experimental results show a 5-fold classification accuracy of 0.85 for the ADNI dataset and 0.70 for the ABIDE dataset. The proposed GABM method has extracted 6 to 65 knowledgeable brain regions in ADNI dataset (and 15 to 75 knowledgeable brain regions in ABIDE dataset). These regions are interpreted as the segments of the brain which are mostly used by the 3D-CNN to extract features for brain disease classification. Experimental results show that besides the model interpretability, the proposed GABM method has increased final performance of the classification model in some cases with respect to model parameters.

Accuracy of MRI Classification Algorithms in a Tertiary Memory Center Clinical Routine Cohort.

BACKGROUND: Automated volumetry software (AVS) has recently become widely available to neuroradiologists. MRI volumetry with AVS may support the diagnosis of dementias by identifying regional atrophy. Moreover, automatic classifiers using machine learning techniques have recently emerged as promising approaches to assist diagnosis. However, the performance of both AVS and automatic classifiers have been evaluated mostly in the artificial setting of research datasets. OBJECTIVE: Our aim was to evaluate the performance of two AVS and an automatic classifier in the clinical routine condition of a memory clinic. METHODS: We studied 239 patients with cognitive troubles from a single memory center cohort. Using clinical routine T1-weighted MRI, we evaluated the classification performance of: 1) univariate volumetry using two AVS (volBrain and Neuroreader™); 2) Support Vector Machine (SVM) automatic classifier, using either the AVS volumes (SVM-AVS), or whole gray matter (SVM-WGM); 3) reading by two neuroradiologists. The performance measure was the balanced diagnostic accuracy. The reference standard was consensus diagnosis by three neurologists using clinical, biological (cerebrospinal fluid) and imaging data and following international criteria. RESULTS: Univariate AVS volumetry provided only moderate accuracies (46% to 71% with hippocampal volume). The accuracy improved when using SVM-AVS classifier (52% to 85%), becoming close to that of SVM-WGM (52 to 90%). Visual classification by neuroradiologists ranged between SVM-AVS and SVM-WGM. CONCLUSION: In the routine practice of a memory clinic, the use of volumetric measures provided by AVS yields only moderate accuracy. Automatic classifiers can improve accuracy and could be a useful tool to assist diagnosis.

Predictive Pattern Classification Can Distinguish Gender Identity Subtypes from Behavior and Brain Imaging.

The exact neurobiological underpinnings of gender identity (i.e., the subjective perception of oneself belonging to a certain gender) still remain unknown. Combining both resting-state functional connectivity and behavioral data, we examined gender identity in cisgender and transgender persons using a data-driven machine learning strategy. Intrinsic functional connectivity and questionnaire data were obtained from cisgender (men/women) and transgender (trans men/trans women) individuals. Machine learning algorithms reliably detected gender identity with high prediction accuracy in each of the four groups based on connectivity signatures alone. The four normative gender groups were classified with accuracies ranging from 48% to 62% (exceeding chance level at 25%). These connectivity-based classification accuracies exceeded those obtained from a widely established behavioral instrument for gender identity. Using canonical correlation analyses, functional brain measurements and questionnaire data were then integrated to delineate nine canonical vectors (i.e., brain-gender axes), providing a multilevel window into the conventional sex dichotomy. Our dimensional gender perspective captures four distinguishable brain phenotypes for gender identity, advocating a biologically grounded reconceptualization of gender dimorphism. We hope to pave the way towards objective, data-driven diagnostic markers for gender identity and transgender, taking into account neurobiological and behavioral differences in an integrative modeling approach.

Interobserver Reliability When Classifying MR Imaging of the Lumbar Spine: Written Instructions Alone Do Not Suffice.

PURPOSE: Numerous classification systems have been proposed to analyze lumbar spine MRI scans. When evaluating these systems, most studies draw their conclusions from measurements of experienced clinicians. The aim of this study was to evaluate the impact of specific measurement training on interobserver reliability in MRI classification of the lumbar spine. METHODS: Various measurement and classification systems were assessed for their interobserver reliability in 30 MRIs from patients with chronic lumbar back and sciatic pain. Two observers were experienced spine surgeons. The third observer was an inexperienced medical student who, prior to the study measurements, in addition to being given the detailed written instructions also given to the surgeons, obtained a list of 20 reference measurements in MRI scans from other patients to practice with. RESULTS: Excellent agreement was observed between the medical student and the spine surgeon who had also created the reference measurements. Between the two spine surgeons, agreement was markedly lower in all systems investigated (e.g., antero-posterior spinal canal diameter intraclass correlation coefficient [ICC] [3.1] = 0.979 vs. ICC [3.1] = 0.857). CONCLUSION: These data warrant the creation of publicly available standardised measurement examples of accepted classification systems to increase reliability of the interpretation of MR images.

Does three-grade classification of T2-weighted increased signal intensity reflect the severity of myelopathy and surgical outcomes in patients with cervical compressive myelopathy? A systematic review and meta-analysis.

The three-grade classification of increased signal intensity (ISI) on T2-weighted magnetic resonance imaging (MRI) is used extensively in patients with cervical compressive myelopathy (CCM). However, the efficacy and value in the prediction of this classification are still unclear and no systematic review and meta-analysis have been conducted on this topic. The objective of this study is to investigate the efficacy and value in prediction of the three-grade classification of ISI on the severity of myelopathy and surgical outcomes. Randomized or non-randomized controlled studies using three-grade classification of ISI (grade 0, none; grade 1, light or obscure; and grade 2, intense or bright) in patients with CCM were sought in the following databases: PubMed, Embase, and Cochrane Library. The pooled Japanese Orthopedic Association (JOA)/modified JOA (mJOA) score, neuro-functional recovery rate, C2-C7 lordotic angle, and range of motion (ROM) were calculated. A total of 8 studies containing 1101 patients were included in this review. Patients in grade 0 had the highest preoperative and postoperative JOA/mJOA score and recovery rate, while those parameters for patients in grade 2 were the lowest. Nevertheless, no statistically significant difference was found regarding the preoperative C2-C7 lordotic angle and ROM among three grades. Our meta-analysis suggests that the three-grade classification of ISI on T2-weighted MRI can reflect the severity of myelopathy and surgical outcomes in patients with CCM. The higher ISI grade indicates more severe myelopathy and surgical outcomes. Overall, the three-grade classification of ISI is instructive and should be used universally.

Fusion of ULS Group Constrained High- and Low-Order Sparse Functional Connectivity Networks for MCI Classification.

Functional connectivity networks, derived from resting-state fMRI data, have been found as effective biomarkers for identifying mild cognitive impairment (MCI) from healthy elderly. However, the traditional functional connectivity network is essentially a low-order network with the assumption that the brain activity is static over the entire scanning period, ignoring temporal variations among the correlations derived from brain region pairs. To overcome this limitation, we proposed a new type of sparse functional connectivity network to precisely describe the relationship of temporal correlations among brain regions. Specifically, instead of using the simple pairwise Pearson's correlation coefficient as connectivity, we first estimate the temporal low-order functional connectivity for each region pair based on an ULS Group constrained-UOLS regression algorithm, where a combination of ultra-least squares (ULS) criterion with a Group constrained topology structure detection algorithm is applied to detect the topology of functional connectivity networks, aided by an Ultra-Orthogonal Least Squares (UOLS) algorithm to estimate connectivity strength. Compared to the classical least squares criterion which only measures the discrepancy between the observed signals and the model prediction function, the ULS criterion takes into consideration the discrepancy between the weak derivatives of the observed signals and the model prediction function and thus avoids the overfitting problem. By using a similar approach, we then estimate the high-order functional connectivity from the low-order connectivity to characterize signal flows among the brain regions. We finally fuse the low-order and the high-order networks using two decision trees for MCI classification. Experimental results demonstrate the effectiveness of the proposed method on MCI classification.