Independent Component Analysis Components Research Articles

Multivariate white matter differences links to cognition in individuals with family history of Alzheimer’s disease and APOE4 genetic risk

AbstractBackgroundGrowing evidence suggests Alzheimer’s disease (AD) develops from a complex cascade of events that vary between individuals. Multivariate approaches have the potential to capture the complexity and heterogeneity of pathologies underlying AD in a more holistic manner compared to univariate approaches.MethodThe MRI data of 105 older adults with a family history of AD from the PREVENT‐AD dataset were analysed. Forty six of these individuals had the APOE4‐3 genotype (greater risk of developing AD), while 59 subjects had the APOE3‐3 genotype (normal risk). Diffusion‐weighted imaging and multi‐echo magnetization transfer, proton density and T1‐weighted data were used to compute several WM metrics (Fig1 b‐c). We computed the voxel‐wise Mahalanobis distance (D2) in WM between APOE4‐3 individuals and a reference group (APOE3‐3) (Fig1). D2 is a multivariate measure that combines several MRI metrics, accounting for their covariance, and yielding a score indicative of the degree of abnormality at each WM voxel (Fig2). Associations between WM D2 and cognitive performance were then investigated (Fig3). Independent component analysis (ICA) was performed on spatial dimensions and linear regression analyses were conducted between D2 in ICA components and scores on items of the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS). We then extracted the contribution of each MRI feature to D2 in the regions where significant associations were found.ResultA significant negative association (R = ‐0.36, p = 0.014) was found between D2 in a component corresponding to the splenium of the corpus callosum (CC) and scores on RBANS‐immediate memory (Fig3), indicating a link between a greater degree of WM abnormality and poorer performance. Proton density (MPM_PD) and longitudinal relaxation rate (MPM_R1), potentially indicative of WM loss, contributed most to D2 in this region.ConclusionOur findings suggest that microstructural alterations in the posterior part of the CC may play a role in memory deficit in older adults who are at greater risk of developing AD. This is in line with other studies that showed a link between CC degeneration and memory impairments in early AD stages.1 1. Qiu et al. (2016). Inter‐hemispheric dysconnectivity in AD and aMCI. Sci.Rep.

Multivariate white matter differences links to cognition in individuals with family history of Alzheimer’s disease and APOE4 genetic risk

AbstractBackgroundGrowing evidence suggests Alzheimer’s disease (AD) develops from a complex cascade of events that vary between individuals. Multivariate approaches have the potential to capture the complexity and heterogeneity of pathologies underlying AD in a more holistic manner compared to univariate approaches.MethodThe MRI data of 105 older adults with a family history of AD from the PREVENT‐AD dataset were analysed. Forty six of these individuals had the APOE4‐3 genotype (greater risk of developing AD), while 59 subjects had the APOE3‐3 genotype (normal risk). Diffusion‐weighted imaging and multi‐echo magnetization transfer, proton density and T1‐weighted data were used to compute several WM metrics (Fig1 b‐c). We computed the voxel‐wise Mahalanobis distance (D2) in WM between APOE4‐3 individuals and a reference group (APOE3‐3) (Fig1). D2 is a multivariate measure that combines several MRI metrics, accounting for their covariance, and yielding a score indicative of the degree of abnormality at each WM voxel (Fig2). Associations between WM D2 and cognitive performance were then investigated (Fig3). Independent component analysis (ICA) was performed on spatial dimensions and linear regression analyses were conducted between D2 in ICA components and scores on items of the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS). We then extracted the contribution of each MRI feature to D2 in the regions where significant associations were found.ResultA significant negative association (R = ‐0.36, p = 0.014) was found between D2 in a component corresponding to the splenium of the corpus callosum (CC) and scores on RBANS‐immediate memory (Fig3), indicating a link between a greater degree of WM abnormality and poorer performance. Proton density (MPM_PD) and longitudinal relaxation rate (MPM_R1), potentially indicative of WM loss, contributed most to D2 in this region.ConclusionOur findings suggest that microstructural alterations in the posterior part of the CC may play a role in memory deficit in older adults who are at greater risk of developing AD. This is in line with other studies that showed a link between CC degeneration and memory impairments in early AD stages.1 1. Qiu et al. (2016). Inter‐hemispheric dysconnectivity in AD and aMCI. Sci.Rep.

Denoising of blasting vibration signals based on CEEMDAN-ICA algorithm.

Monitoring of blasting vibration signals can make the collected blasting signals noisy due to various factors such as on-site actual construction conditions, equipment, and instruments. Thus, the acquired signals should be preprocessed before analyzing the blasting vibration signals. The current study proposes a blasting vibration denoising method based on CEEMDAN-ICA to alleviate the noise component in the blasting signals effectively. The collected signal is first decomposed through the CEMMDAN algorithm to extract the IMF components of different frequency bands. Next, the collected signal is estimated using the ICA algorithm to attain corresponding ICA components. Finally, the arrangement entropy of the ICA components is calculated for signal reconstruction to attain a small noise blasting vibration signal. Simulations are performed to evaluate the feasibility of the presented algorithm and compare its efficiency with the traditional algorithms. The results demonstrate that this algorithm has specific advantages over other algorithms, which can more accurately denoise the original signal and retain the effective signals, providing a new denoising method for subsequent signal analysis.

Independent component analysis (ICA) applied to dynamic oxygen-enhanced MRI (OE-MRI) for robust functional lung imaging at 3 T.

Dynamic lung oxygen-enhanced MRI (OE-MRI) is challenging due to the presence of confounding signals and poor signal-to-noise ratio, particularly at 3 T. We have created a robust pipeline utilizing independent component analysis (ICA) to automatically extract the oxygen-induced signal change from confounding factors to improve the accuracy and sensitivity of lung OE-MRI. Dynamic OE-MRI was performed on healthy participants using a dual-echo multi-slice spoiled gradient echo sequence at 3 T and cyclical gas delivery. ICA was applied to each echo within a thoracic mask. The ICA component relating to the oxygen-enhancement signal was automatically identified using correlation analysis. The oxygen-enhancement component was reconstructed, and the percentage signal enhancement (PSE) was calculated. The lung PSE of current smokers was compared with nonsmokers; scan-rescan repeatability, ICA pipeline repeatability, and reproducibility between two vendors were assessed. ICA successfully extracted a consistent oxygen-enhancement component for all participants. Lung tissue and oxygenated blood displayed the opposite oxygen-induced signal enhancements. A significant difference in PSE was observed between the lungs of current smokers and nonsmokers. The scan-rescan repeatability and the ICA pipeline repeatability were good. The developed pipeline demonstrated sensitivity to the signal enhancements of the lung tissue and oxygenated blood at 3 T. The difference in lung PSE between current smokers and nonsmokers indicates a likely sensitivity to lung function alterations that may be seen in mild pathology, supporting future use of our methods in patient studies.

An inclusive paradigm to study mu-rhythm properties

Studying mu-rhythm in developmental disorders is crucial for identifying the origin of motor and social malfunctioning. However, the commonly used mu-rhythm experimental protocol, that requires following instructions, is challenging for children with motor and cognitive deficits. Here we present an inclusive experimental procedure that contains passive hand movement, closed and open eyes and hand movement observation and examine properties of EEG mu-rhythm obtained in this paradigm in 51 typically developing children and adults aged from 4 to 31 years. The independent component analysis (ICA) was used to separate occipital alpha- and mu-rhythm components and showed better performance than the channel-wise analysis. The identified mu-rhythm ICA components were localized above the left and right sensorimotor cortex, demonstrating suppression both to passive hand movement and hand movement observation and no difference in power between closed- and open-eye conditions. No interhemispheric differences were observed. The alpha-rhythm ICA components were localized in occipital regions and demonstrate characteristic suppression to open-eye conditions. The mu-rhythm frequency of peak suppression to passive hand movement as well as the amount of suppression increased with age. The beta-band mu-rhythm activity, while being less pronounced, was also suppressed both during passive hand movement and hand movement observation, while to a lesser degree than alpha-band mu during passive hand movement. Thus, we confirmed the classical properties of mu-rhythm and for the first time showed the developmental trajectory of mu-rhythm properties obtained during passive hand movement. The proposed experimental protocol and pipeline can be used further in studies of the mu-rhythm in challenging populations.

Development and evaluation of 5FUCore, OxaliCore, and IrinoCore transcriptomic signatures for predicting response to individual drugs in the FOLFIRINOX regimen: Towards toxicity reduction and personalized chemotherapy in PDAC.

4141 Background: Recently, we have successfully developed a transcriptomic signature that predicts the response to gemcitabine treatment in patients with localized pancreatic ductal adenocarcinoma (PDAC) (PMID: 36178036) and also in those with metastatic PDAC (PMID: 36496056). This was achieved by using an approach that combined several preclinical PDAC models. The FOLFIRINOX regimen (FFX) consisting of 5-Fluorouracil (5FU), Oxaliplatin, and Irinotecan is the more effective treatment option, but the high toxicity limits its use. In this study, we developed and analyzed three different transcriptomic signatures for the drugs composing FFX separately and evaluated their clinical utility. Methods: We combined the transcriptomic profile of 14 patients derived xenograft (PDX) and 31 patients derived cell cultures (PDC) with their corresponding chemotherapy response profiles. We developed predictive signatures for 5FU ( 5FUCore), Oxaliplatin ( OxaliCore), and Irinotecan ( IrinoCore), as previously reported for the GemCore signature (PMID: 36178036). They were then validated in three independent cohorts of patients: COMPASS (n = 94), Linehan (n = 28), and a new cohort of 87 advanced PDAC patients. Results: Three ICA components that reflect the chemotherapy response profiles for 5FU, oxaliplatin, and irinotecan were evaluated in the COMPASS cohort to determine their ability to identify sensitive patients in the FFX arm. Patients predicted as sensitive to each drug separately had a median overall survival (OS) of 77.6 (95% CI, 67.9-NR, P =0.02), 78.7 (95% CI, 57.6-NR, P= 0.002), and 70.4 (95% CI, 57.6-NR, P= 0.02) months for 5FU, oxaliplatin, and irinotecan, respectively. Notably, the model that integrated the group of sensitive patients to two or three drugs had the highest median OS in the COMPASS and Linehan cohorts with a HR of 0.25 (95% CI, 0.11-0.57; P < 0.001) and 0.11 (95% CI, 0.02-0.70; P= 0.019), respectively. Importantly, they also show a better objective response than the other groups ( P= 0.009). In the advanced PDAC group, the signatures showed a strong ability to discriminate sensitive patients in the FFX arm with a median OS of 20.1 (95% CI, 2.3-NR, P= 0.006), 13.64 (95% CI, 9.57-NR, P= 0.003), and 15.87 (95% CI, 2.9-NR, P= 0.002) months for 5FU, oxaliplatin, and irinotecan, respectively. Finally, we evaluated the best combination of two drugs in the pooled group. Patients sensitive to both 5FU and oxaliplatin showed the best response with 60% of patients achieving partial response. Conclusions: The signatures developed provide a tool for predicting responsiveness to the individual therapeutic drugs of the FFX regimen. They may ultimately lead to the development of a personalized strategy for combined chemotherapy regimens, increasing efficacy and reducing toxicity in PDAC.

Identifying reproducible resting state networks and functional connectivity alterations following chronic restraint stress in anaesthetized rats.

Resting-state functional MRI (rs-fMRI) in rodent models have the potential to bridge invasive experiments and observational human studies, increasing our understanding of functional alterations in the brains of patients with depression. A major limitation in current rodent rs-fMRI studies is that there has been no consensus on healthy baseline resting-state networks (RSNs) that are reproducible in rodents. Therefore, the present study aimed to construct reproducible RSNs in a large dataset of healthy rats and then evaluate functional connectivity changes within and between these RSNs following a chronic restraint stress (CRS) model within the same animals. A combined MRI dataset of 109 Sprague Dawley rats at baseline and after two weeks of CRS, collected during four separate experiments conducted by our lab in 2019 and 2020, was re-analysed. The mICA and gRAICAR toolbox were first applied to detect optimal and reproducible ICA components and then a hierarchical clustering algorithm (FSLNets) was applied to construct reproducible RSNs. Ridge-regularized partial correlation (FSLNets) was used to evaluate the changes in the direct connection between and within identified networks in the same animals following CRS. Four large-scale networks in anesthetised rats were identified: the DMN-like, spatial attention-limbic, corpus striatum, and autonomic network, which are homologous across species. CRS decreased the anticorrelation between DMN-like and autonomic network. CRS decreased the correlation between amygdala and a functional complex (nucleus accumbens and ventral pallidum) in the right hemisphere within the corpus striatum network. However, a high individual variability in the functional connectivity before and after CRS within RSNs was observed. The functional connectivity changes detected in rodents following CRS differ from reported functional connectivity alterations in patients with depression. A simple interpretation of this difference is that the rodent response to CRS does not reflect the complexity of depression as it is experienced by humans. Nonetheless, the high inter-subject variability of functional connectivity within networks suggests that rats demonstrate different neural phenotypes, like humans. Therefore, future efforts in classifying neural phenotypes in rodents might improve the sensitivity and translational impact of models used to address aetiology and treatment of psychiatric conditions including depression.

Extracting electromyographic signals from multi-channel LFPs using independent component analysis without direct muscular recording

Electromyography (EMG) has been commonly used for the precise identification of animal behavior. However, it is often not recorded together with invivo electrophysiology due to the need for additional surgeries and setups and the high risk of mechanical wire disconnection. While independent component analysis (ICA) has been used to reduce noise from field potential data, there has been no attempt to proactively use the removed "noise," of which EMG signals are thought to be one of the major sources. Here, we demonstrate that EMG signals can be reconstructed without direct EMG recording using the "noise" ICA component from local field potentials. The extracted component is highly correlated with directly measured EMG, termed IC-EMG. IC-EMG is useful for measuring an animal's sleep/wake, freezing response, and non-rapid eye movement (NREM)/REM sleep states consistently with actual EMG. Our method has advantages in precise and long-termbehavioral measurement in wide-ranging invivo electrophysiology experiments.

The Influence Assessment of Artifact Subspace Reconstruction on the EEG Signal Characteristics

EEG signals may be affected by physiological and non-physiological artifacts hindering the analysis of brain activity. Blind source separation methods such as independent component analysis (ICA) are effective ways of improving signal quality by removing components representing non-brain activity. However, most ICA-based artifact removal strategies have limitations, such as individual differences in visual assessment of components. These limitations might be reduced by introducing automatic selection methods for ICA components. On the other hand, new fully automatic artifact removal methods are developed. One of such method is artifact subspace reconstruction (ASR). ASR is a component-based approach, which can be used automatically and with small calculation requirements. The ASR was originally designed to be run not instead of, but in addition to ICA. We compared two automatic signal quality correction approaches: the approach based only on ICA method and the approach where ASR was applied additionally to ICA and run before the ICA. The case study was based on the analysis of data collected from 10 subjects performing four popular experimental paradigms, including resting-state, visual stimulation and oddball task. Statistical analysis of the signal-to-noise ratio showed a significant difference, but not between ICA and ASR followed by ICA. The results show that both methods provided a signal of similar quality, but they were characterised by different usabilities.

Pre-surgical features of intrinsic brain networks predict single and joint epilepsy surgery outcomes

Despite the effectiveness of surgical interventions for the treatment of intractable focal temporal lobe epilepsy (TLE), the substrates that support good outcomes are poorly understood. While algorithms have been developed for the prediction of either seizure or cognitive/psychiatric outcomes alone, no study has reported on the functional and structural architecture that supports joint outcomes. We measured key aspects of pre-surgical whole brain functional/structural network architecture and evaluated their ability to predict post-operative seizure control in combination with cognitive/psychiatric outcomes. Pre-surgically, we identified the intrinsic connectivity networks (ICNs) unique to each person through independent component analysis (ICA), and computed: (1) the spatial–temporal match between each person’s ICA components and established, canonical ICNs, (2) the connectivity strength within each identified person-specific ICN, (3) the gray matter (GM) volume underlying the person-specific ICNs, and (4) the amount of variance not explained by the canonical ICNs for each person. Post-surgical seizure control and reliable change indices of change (for language [naming, phonemic fluency], verbal episodic memory, and depression) served as binary outcome responses in random forest (RF) models. The above functional and structural measures served as input predictors. Our empirically derived ICN-based measures customized to the individual showed that good joint seizure and cognitive/psychiatric outcomes depended upon higher levels of brain reserve (GM volume) in specific networks. In contrast, singular outcomes relied on systematic, idiosyncratic variance in the case of seizure control, and the weakened pre-surgical presence of functional ICNs that encompassed the ictal temporal lobe in the case of cognitive/psychiatric outcomes. Our data made clear that the ICNs differed in their propensity to provide reserve for adaptive outcomes, with some providing structural (brain), and others functional (cognitive) reserve. Our customized methodology demonstrated that when substantial unique, patient-specific ICNs are present prior to surgery there is a reliable association with poor post-surgical seizure control. These ICNs are idiosyncratic in that they did not match the canonical, normative ICNs and, therefore, could not be defined functionally, with their location likely varying by patient. This important finding suggested the level of highly individualized ICN’s in the epileptic brain may signal the emergence of epileptogenic activity after surgery.

Clinical and cortical similarities identified between bipolar disorder I and schizophrenia: A multivariate approach.

Structural neuroimaging studies have identified similarities in the brains of individuals diagnosed with schizophrenia (SZ) and bipolar I disorder (BP), with overlap in regions of gray matter (GM) deficits between the two disorders. Recent studies have also shown that the symptom phenotypes associated with SZ and BP may allow for a more precise categorization than the current diagnostic criteria. In this study, we sought to identify GM alterations that were unique to each disorder and whether those alterations were also related to unique symptom profiles. We analyzed the GM patterns and clinical symptom presentations using independent component analysis (ICA), hierarchical clustering, and n-way biclustering in a large (N ∼ 3,000), merged dataset of neuroimaging data from healthy volunteers (HV), and individuals with either SZ or BP. Component A showed a SZ and BP < HV GM pattern in the bilateral insula and cingulate gyrus. Component B showed a SZ and BP < HV GM pattern in the cerebellum and vermis. There were no significant differences between diagnostic groups in these components. Component C showed a SZ < HV and BP GM pattern bilaterally in the temporal poles. Hierarchical clustering of the PANSS scores and the ICA components did not yield new subgroups. N-way biclustering identified three unique subgroups of individuals within the sample that mapped onto different combinations of ICA components and symptom profiles categorized by the PANSS but no distinct diagnostic group differences. These multivariate results show that diagnostic boundaries are not clearly related to structural differences or distinct symptom profiles. Our findings add support that (1) BP tend to have less severe symptom profiles when compared to SZ on the PANSS without a clear distinction, and (2) all the gray matter alterations follow the pattern of SZ < BP < HV without a clear distinction between SZ and BP.

Data-driven approach for the delineation of the irritative zone in epilepsy in MEG.

The reliable identification of the irritative zone (IZ) is a prerequisite for the correct clinical evaluation of medically refractory patients affected by epilepsy. Given the complexity of MEG data, visual analysis of epileptiform neurophysiological activity is highly time consuming and might leave clinically relevant information undetected. We recorded and analyzed the interictal activity from seven patients affected by epilepsy (Vectorview Neuromag), who successfully underwent epilepsy surgery (Engel > = II). We visually marked and localized characteristic epileptiform activity (VIS). We implemented a two-stage pipeline for the detection of interictal spikes and the delineation of the IZ. First, we detected candidate events from peaky ICA components, and then clustered events around spatio-temporal patterns identified by convolutional sparse coding. We used the average of clustered events to create IZ maps computed at the amplitude peak (PEAK), and at the 50% of the peak ascending slope (SLOPE). We validated our approach by computing the distance of the estimated IZ (VIS, SLOPE and PEAK) from the border of the surgically resected area (RA). We identified 25 spatiotemporal patterns mimicking the underlying interictal activity (3.6 clusters/patient). Each cluster was populated on average by 22.1 [15.0-31.0] spikes. The predicted IZ maps had an average distance from the resection margin of 8.4 ± 9.3 mm for visual analysis, 12.0 ± 16.5 mm for SLOPE and 22.7 ±. 16.4 mm for PEAK. The consideration of the source spread at the ascending slope provided an IZ closer to RA and resembled the analysis of an expert observer. We validated here the performance of a data-driven approach for the automated detection of interictal spikes and delineation of the IZ. This computational framework provides the basis for reproducible and bias-free analysis of MEG recordings in epilepsy.

Automatic Transmission Bearing Fault Diagnosis Based on Comprehensive Index Method and Convolutional Neural Network

Rolling-element bearing fault diagnosis has some problems in the applied environment, such as low signal-to-noise ratio, weak feature extraction, low efficiency of feature learning and the complex structure of diagnosis models. A fault diagnosis method based on the comprehensive index method, complete ensemble empirical mode decomposition with adaptive noise independent component analysis (CEEMDANICA) and two-dimensional convolutional neural network (TDCNN) is proposed. Firstly, the original vibration signal of the bearing is preprocessed by CEEMDANICA, and the ICA components with different frequencies are obtained. Secondly, the ICA components are selected as the sample set by using multiscale permutation entropy, correlation coefficient, kurtosis and box dimension. Finally, the sample set are trained and tested by a DCNN model to realize the fault diagnosis of different bearing fault types. In order to verify the reliability of the method, a bearing fault vibration monitoring platform for an electric vehicle two-speed automatic transmission was built to collect the bearing vibration signals of multiple fault types under different working conditions. The diagnostic accuracy of several deep learning models is compared. The results show that the proposed method can realize the single and compound fault diagnosis of rolling-element bearings in an automatic transmission, with a high degree of accuracy.

Resting-state functional MRI for motor cortex mapping in childhood-onset focal epilepsy.

Task-based functional MRI (fMRI) mapping of the motor function prior to epilepsy surgery has limitations in children with epilepsy. We present a data-driven method to automatically delineate the motor cortex using task-free, resting-state fMRI (rsfMRI) data. We used whole-brain rsfMRI for independent component analysis (ICA). A template matching process with Discriminability Index-based Component Identification score was used for each participant to select and combine motor ICA components in their native brain space, resulting in a whole-brain ICA Motor Map (wIMM). We validated wIMM by comparing individual results with bilateral finger-tapping motor task fMRI activation, and evaluated its reproducibility in controls. Data from 64 patients and 12 controls were used to generate group wIMM maps. The hit rate between wIMM and motor task activation ranged from 60% to 79% across all participants. Sensitivity of wIMM for capturing the task activation peak was 87.5% among 32 patients and 100% in 12 controls with available motor task results. We also showed high similarity in repeated runs in controls. Our results show the sensitivity and reproducibility of an automated motor mapping method based on ICA analysis of rsfMRI in children with epilepsy. The ICA maps may provide different, but useful, information than task fMRI. Future studies will expand our method to mapping other brain functions, and may lead to a surgical planning tool for patients who cannot perform task fMRI and help predict their postsurgical function.

The optimized combination of aCompCor and ICA-AROMA to reduce motion and physiologic noise in task fMRI data

One of the main challenges in fMRI processing is filtering the task BOLD signals from the noise. Independent component analysis with automatic removal of motion artifacts (ICA-AROMA) reduces motion artifacts by identifying ICA noise components based on their location at the brain edges and cerebrospinal fluid (CSF), high frequency content and correlation with motion regressors. In anatomical component correction (aCompCor), physiological noise regressors extracted from CSF were regressed out from the fMRI time series. In this study, we compared three methods to combine aCompCor and ICA-AROMA denoising in one denoising step. In the first analysis, we regressed the temporal signals of the ICA components identified as noise by ICA-AROMA together with the noise signals determined by aCompCor from the fMRI signals. For the second and third analyses, the correlation between the temporal signals of the ICA components and the aCompCor noise signals was used as an additional criterion to identify the noise components. In the second analysis, the temporal signals of the ICA components classified as noise were regressed from the fMRI signals. In the third analysis, the noise components were removed. To compare the denoising strategies, we examined the fractional amplitude of low-frequency fluctuations (fALFF) and the overlap between the contrast maps. Our results revealed that including the aCompCor noise signals as regressors in ICA-AROMA resulted in more correctly identified noise components, higher fALFF values, and larger activation maps. Moreover, combining the temporal signals of the noise components identified by ICA-AROMA with the aCompCor signals in a noise regression matrix resulted in deactivations. These results suggest that using the correlation between the ICA component temporal signals and the aCompCor signals as noise identification criteria in ICA-AROMA is the best approach for combining both denoising methods.

455-P: Artificial Intelligence Approach to Treatment Classification in Painful Diabetic Neuropathy

Aim: Our study predicted treatment responses in patients with painful DPN (diabetic peripheral neuropathy) by developing a deep learning model using resting state functional magnetic resonance imaging (fMRI) neuroimaging datasets. Methods: Forty-three consecutive patients who received intravenous lidocaine treatment for painful DPN were assessed. All subjects (responders n=29 and non-responders n=14) underwent detailed clinical and neurophysiological assessments to phenotype their pain sensory profile. Subjects also underwent brain resting-state fMRI. After pre-processing we performed a group concatenated independent component analysis (ICA) set to 30 components and automatically chose 7 highly correlated (p&amp;lt;0.05) ICA components. A 3D convolutional neural network (CNN) classification framework was trained using a VoxNet based architecture. This deep learning architecture compared models using (1) 7 correlated ICA networks 2) all 30 ICA networks generated 3) pre-processed resting state images. Results: The deep learning treatment response classification in a ten-fold cross validation experiment using 7 ICA spatial maps has a mean AUC of 0.85 and an F1-Score of 0.90. However, with the extra information of all 30 ICA maps the mean AUROC increased to 0.97 with an F1-Score of 0.95. Using only pre-processed resting-state fMRI data achieved suboptimal F1-Score of 69% and AUC score of 44%. Conclusion: Through the use of our deep learning model, we have demonstrated high classification performance. Our method improves painful DPN treatment efficiency by stratifying patients to receive the correct treatment from the outset. We believe to our knowledge this is the first study utilising deep learning methods to classify treatment response in painful DPN. Disclosure K.Teh: None. I.D.Wilkinson: None. G.P.Sloan: None. S.Tesfaye: Advisory Panel; Astellas Pharma Inc., Bayer AG, Grünenthal Group, Nevro Corp., Wörwag Pharma GmbH &amp; Co. KG, Speaker's Bureau; Eva Pharma, Pfizer Inc., Viatris Inc. D.Selvarajah: None. Funding National Instituteof Health Research Efficacy and Mechanism Evaluation Programme (NIHR 129921) European Foundation for the Study of Diabetes (Microvascular Complications Project Grant) University of Sheffield, Health Education England, Knowledge Exchange

Automated Classification of Resting-State fMRI ICA Components Using a Deep Siamese Network.

Manual classification of functional resting state networks (RSNs) derived from Independent Component Analysis (ICA) decomposition can be labor intensive and requires expertise, particularly in large multi-subject analyses. Hence, a fully automatic algorithm that can reliably classify these RSNs is desirable. In this paper, we present a deep learning approach based on a Siamese Network to learn a discriminative feature representation for single-subject ICA component classification. Advantages of this supervised framework are that it requires relatively few training data examples and it does not require the number of ICA components to be specified. In addition, our approach permits one-shot learning, which allows generalization to new classes not seen in the training set with only one example of each new class. The proposed method is shown to out-perform traditional convolutional neural network (CNN) and template matching methods in identifying eleven subject-specific RSNs, achieving 100% accuracy on a holdout data set and over 99% accuracy on an outside data set. We also demonstrate that the method is robust to scan-rescan variation. Finally, we show that the functional connectivity of default mode and salience networks identified by the proposed technique is altered in a group analysis of mild traumatic brain injury (TBI), severe TBI, and healthy subjects.

Structural Covariance Network as an Endophenotype in Alzheimer's Disease-Susceptible Single-Nucleotide Polymorphisms and the Correlations With Cognitive Outcomes.

The cognitive manifestations of Alzheimer’s disease (AD) are related to brain network degeneration, and genetic differences may mediate network degeneration. Several AD-susceptible loci have been reported to involve amyloid or tau cascades; however, their relationships with gray matter (GM) volume and cognitive outcomes have yet to be established. We hypothesized that single-nucleotide polymorphism genotype groups may interact with apolipoprotein E4 (ApoE4) status or independently exert an effect on cognitive outcomes. We also hypothesized that GM structural covariance networks (SCNs) may serve as an endophenotype of the genetic effect, which, in turn, may be related to neurobehavior test scores. Gray matter SCNs were constructed in 324 patients with AD using T1 magnetic resonance imaging with independent component analysis (ICA). We assessed the effects of 15 genetic loci (rs9349407, rs3865444, rs670139, rs744373, rs3851179, rs11136000, rs3764650, rs610932, rs6887649, rs7849530, rs4866650, rs3765728, rs34011, rs6656401, and rs597668) using additive, recessive, and dominant models on cognitive outcomes. Statistical analysis was performed to explore the independent role of each locus, interactions with ApoE4 status, and relationships to GM ICA network intensity score. For outcome measures, we used the Mini-Mental State Examination (MMSE), Cognitive Abilities Screening Instrument (CASI) total score, and short-term memory (STM) subscores, adjusted for the covariates of education, disease duration, and age. Clinically, the CD2AP G allele showed a protective role in MMSE, CASI total, and CASI-STM scores independently or via interactions with non-ApoE4 status, while the CR1 A genotype group was associated with lower STM subscores independent of ApoE4 status. Three loci showed synergic interactions with ApoE4: BIN 1, MS4A6A, and FTMT. Of the 15 meaningful ICA components, 5 SCNs (anterior and posterior hippocampus, right temporal, left thalamus, default mode network) showed relationships with general cognitive performance, in which only the ApoE4 and MS4A6A genotype groups were independently related to the hippocampus network. The genetic loci MS4A6A, BIN1, CLU, CR1, BIN1, PICALM, and FGF1 influenced the networks independently or in synergy. This study suggests that AD-susceptible loci may each exert clinical significance independently through interactions with ApoE4 status or through SCNs as an endophenotype and that this effect is associated with the cognitive outcomes.

Automatic classification of ICA components from infant EEG using MARA

Automated systems for identifying and removing non-neural ICA components are growing in popularity among EEG researchers of adult populations. Infant EEG data differs in many ways from adult EEG data, but there exists almost no specific system for automated classification of source components from paediatric populations. Here, we adapt one of the most popular systems for adult ICA component classification for use with infant EEG data. Our adapted classifier significantly outperformed the original adult classifier on samples of naturalistic free play EEG data recorded from 10 to 12-month-old infants, achieving agreement rates with the manual classification of over 75% across two validation studies (n = 44, n = 25). Additionally, we examined both classifiers’ ability to remove stereotyped ocular artifact from a basic visual processing ERP dataset compared to manual ICA data cleaning. Here, the new classifier performed on level with expert manual cleaning and was again significantly better than the adult classifier at removing artifact whilst retaining a greater amount of genuine neural signal operationalised through comparing ERP activations in time and space. Our new system (iMARA) offers developmental EEG researchers a flexible tool for automatic identification and removal of artifactual ICA components.

P 40. β-γ phase-amplitude coupling during movements in Parkinson's disease

Question. Enhanced phase amplitude coupling (PAC) between β (13–30 Hz) and broadband γ (50–150 Hz) activities has been found in surface EEG recordings from Parkinson patients at rest [1]. However, the dynamics of PAC during voluntary movement and its relationship with motor impairment of Parkinson“s disease (PD) remain unclear. Methods. Experimental design : We recorded 64-channel EEG signals from 19 PD patients and 20 sex- and age-matched healthy controls at rest, during a repetitive pressing task and during repetitive index finger tapping performed at the patients“ own pace and at their fastest pace. Source analysis : We obtained source signals by projecting EEG sensor signals to the individual cortical surfaces. Independent component analysis (ICA) was applied for separating temporally independent subnetworks of each task in each of the 4 brain regions: premotor cortex, primary motor cortex, primary somatosensory cortex and primary somatosensory complex. PAC analysis: Pairwise transient PAC values were calculated among ICA components of each region by zscore-normalized mean vector length [2] in 300ms windows shifted with 50ms time steps in 3s epochs which included an interval of 1s before, and 2s after the onset of the movement. Statistics : We applied a non-parametric three-way ANOVA [3] to test the effects of the disease state, tasks and different brain regions on PAC. For post-hoc testing, a two tailed Wilcoxon ranksum test was used between groups and Wilcoxon signrank test within groups. Results. As expected, maximum tapping rate was lower, and fastest movements were executed more irregularly in patients compared to controls. ANOVA performed on the β-γ PAC values averaged over the entire 3s-epoch revealed an interaction of disease state and tasks ( p < 0.001). Post-hoc testing showed enhanced PAC in patients compared with controls in the resting state ( p = 0.018), in agreement with previous findings, but not in any of the tapping tasks. However, PAC during the pressing task remained elevated in the patients compared to controls ( p = 0.020). Analysis of the temporal dynamics revealed reduction of PAC after the onset of the movement (0–0.25 s) compared with that of the pre-onset period (−0.5 to −0.25 s). The reduction of PAC relative to pre-onset period correlated with the desynchronization in the β power. However, the PAC reduction relative to the resting state was not correlated with the β power reduction. Discussion. Persistence of PAC differences during the pressing task on one hand, and disappearance of PAC differences during the tapping tasks on the other hand, suggest that exaggerated β-broadband γ PAC, as detected in the resting EEG, is not necessarily directly related to impaired motor behavior in PD patients. As our previous findings [1] have shown that abnormal PAC at rest is unrelated to the β band power, partial renormalization of PAC by movement-related β band desynchronization may suggest that mechanisms modulating β activity during movement may selectively affect β activity involved in abnormal coupling.