Increasing the reliability of data analysis of functional magnetic resonance imaging by applying a new blockwise permutation method

Three dimensional radial echo planar imaging for functional MRI

PURPOSE

To demonstrate a novel 3D radial echo planar imaging (3D REPI) sequence for flexible, rapid, and motion-robust sampling in fMRI.

METHODS

The 3D REPI method expands on the recently described golden angle rotated EPI trajectory using radial batched internal navigator echoes (TURBINE) approach by exploiting the unused perpendicular direction in the EPI readout to form fast analogues of rotated stack of stars or spirals trajectories that cover all 3 dimensions of k-space. An iterative conjugate gradient algorithm with SENSE reconstruction and time-segmented non-uniform fast Fourier transform (FFT) was used for parallel imaging acceleration and to account for the effects of B₀ inhomogeneity. The golden angle rotation allowed for sliding window reconstruction schemes to be applied in brain BOLD fMRI experiments.

RESULTS

Combined whole brain visual and motor fMRI experiments were successfully carried out on a clinical 3T scanner at 2 mm isotropic and 1 × 1 × 2 mm³ resolutions using the 3D REPI design. Improved sampling characteristics and image quality were observed for twisted trajectories at the expense of prolonged readout times and off-resonance effects. The ability to correct for rigid motion correction was also demonstrated.

CONCLUSIONS

3D REPI presents a flexible approach for segmented volumetric fMRI with motion correction and high in-plane spatial resolutions. Improved BOLD fMRI brain activation maps were obtained using a sliding window reconstruction.

A retrospective physiological noise correction method for oscillating steady-state imaging

PURPOSE

Oscillating steady-state imaging (OSSI) is an SNR-efficient steady-state sequence with T 2 ∗ sensitivity suitable for FMRI. Due to the frequency sensitivity of the signal, respiration- and drift-induced field changes can create unwanted signal fluctuations. This study aims to address this issue by developing retrospective signal correction methods that utilize OSSI signal properties to denoise task-based OSSI FMRI experiments.

METHODS

A retrospective denoising approach was developed that leverages the unique signal properties of OSSI to perform denoising without a manually specified noise region of interest and works with both voxel timecourses (oscillating steady-state correction [OSSCOR]) or FID timecourses (F-OSSCOR). Simulations were performed to estimate the number of principal components optimal for denoising. In vivo experiments at 3 T field strength were conducted to compare the performance of proposed methods against a standard principal component analysis-based method, measured using mean t score within an region of interest, number of activations, and mean temporal SNR.

RESULTS

Correction using OSSCOR was significantly better than the standard method in all metrics. Correction using F-OSSCOR was not significantly different from the standard method using an equal number of principal components. Increasing the number of OSSCOR principal components decreased activation strength and increased the number of suspected false positives. However, increasing the number of principal components in F-OSSCOR increased activation strength with little to no increase in false activation.

CONCLUSION

Both OSSCOR and F-OSSCOR substantially reduce physiological noise components and increase temporal SNR, improving the functional results of task-based OSSI functional experiments. F-OSSCOR demonstrates a proof of concept utilization of coil-localized FID signal information for physiological noise correction.

Bagging improves reproducibility of functional parcellation of the human brain

Increasing the reproducibility of neuroimaging measurement addresses a central impediment to the advancement of human neuroscience and its clinical applications. Recent efforts demonstrating variance in functional brain organization within and between individuals shows a need for improving reproducibility of functional parcellations without long scan times. We apply bootstrap aggregation, or bagging, to the problem of improving reproducibility in functional parcellation. We use two large datasets to demonstrate that compared to a standard clustering framework, bagging improves the reproducibility and test-retest reliability of both cortical and subcortical functional parcellations across a range of sites, scanners, samples, scan lengths, clustering algorithms, and clustering parameters (e.g., number of clusters, spatial constraints). With as little as 6 ​min of scan time, bagging creates more reproducible group and individual level parcellations than standard approaches with twice as much data. This suggests that regardless of the specific parcellation strategy employed, bagging may be a key method for improving functional parcellation and bringing functional neuroimaging-based measurement closer to clinical impact.

Connecting concepts in the brain by mapping cortical representations of semantic relations

In the brain, the semantic system is thought to store concepts. However, little is known about how it connects different concepts and infers semantic relations. To address this question, we collected hours of functional magnetic resonance imaging data from human subjects listening to natural stories. We developed a predictive model of the voxel-wise response and further applied it to thousands of new words. Our results suggest that both semantic categories and relations are represented by spatially overlapping cortical patterns, instead of anatomically segregated regions. Semantic relations that reflect conceptual progression from concreteness to abstractness are represented by cortical patterns of activation in the default mode network and deactivation in the frontoparietal attention network. We conclude that the human brain uses distributed networks to encode not only concepts but also relationships between concepts. In particular, the default mode network plays a central role in semantic processing for abstraction of concepts.

Variational autoencoder: An unsupervised model for encoding and decoding fMRI activity in visual cortex

Goal-driven and feedforward-only convolutional neural networks (CNN) have been shown to be able to predict and decode cortical responses to natural images or videos. Here, we explored an alternative deep neural network, variational auto-encoder (VAE), as a computational model of the visual cortex. We trained a VAE with a five-layer encoder and a five-layer decoder to learn visual representations from a diverse set of unlabeled images. Using the trained VAE, we predicted and decoded cortical activity observed with functional magnetic resonance imaging (fMRI) from three human subjects passively watching natural videos. Compared to CNN, VAE could predict the video-evoked cortical responses with comparable accuracy in early visual areas, but relatively lower accuracy in higher-order visual areas. The distinction between CNN and VAE in terms of encoding performance was primarily attributed to their different learning objectives, rather than their different model architecture or number of parameters. Despite lower encoding accuracies, VAE offered a more convenient strategy for decoding the fMRI activity to reconstruct the video input, by first converting the fMRI activity to the VAE's latent variables, and then converting the latent variables to the reconstructed video frames through the VAE's decoder. This strategy was more advantageous than alternative decoding methods, e.g. partial least squares regression, for being able to reconstruct both the spatial structure and color of the visual input. Such findings highlight VAE as an unsupervised model for learning visual representation, as well as its potential and limitations for explaining cortical responses and reconstructing naturalistic and diverse visual experiences.

Neural Encoding and Decoding with Deep Learning for Dynamic Natural Vision

Convolutional neural network (CNN) driven by image recognition has been shown to be able to explain cortical responses to static pictures at ventral-stream areas. Here, we further showed that such CNN could reliably predict and decode functional magnetic resonance imaging data from humans watching natural movies, despite its lack of any mechanism to account for temporal dynamics or feedback processing. Using separate data, encoding and decoding models were developed and evaluated for describing the bi-directional relationships between the CNN and the brain. Through the encoding models, the CNN-predicted areas covered not only the ventral stream, but also the dorsal stream, albeit to a lesser degree; single-voxel response was visualized as the specific pixel pattern that drove the response, revealing the distinct representation of individual cortical location; cortical activation was synthesized from natural images with high-throughput to map category representation, contrast, and selectivity. Through the decoding models, fMRI signals were directly decoded to estimate the feature representations in both visual and semantic spaces, for direct visual reconstruction and semantic categorization, respectively. These results corroborate, generalize, and extend previous findings, and highlight the value of using deep learning, as an all-in-one model of the visual cortex, to understand and decode natural vision.

LISA improves statistical analysis for fMRI

One of the principal goals in functional magnetic resonance imaging (fMRI) is the detection of local activation in the human brain. However, lack of statistical power and inflated false positive rates have recently been identified as major problems in this regard. Here, we propose a non-parametric and threshold-free framework called LISA to address this demand. It uses a non-linear filter for incorporating spatial context without sacrificing spatial precision. Multiple comparison correction is achieved by controlling the false discovery rate in the filtered maps. Compared to widely used other methods, it shows a boost in statistical power and allows to find small activation areas that have previously evaded detection. The spatial sensitivity of LISA makes it especially suitable for the analysis of high-resolution fMRI data acquired at ultrahigh field (≥7 Tesla).

Transferring and generalizing deep-learning-based neural encoding models across subjects

Recent studies have shown the value of using deep learning models for mapping and characterizing how the brain represents and organizes information for natural vision. However, modeling the relationship between deep learning models and the brain (or encoding models), requires measuring cortical responses to large and diverse sets of natural visual stimuli from single subjects. This requirement limits prior studies to few subjects, making it difficult to generalize findings across subjects or for a population. In this study, we developed new methods to transfer and generalize encoding models across subjects. To train encoding models specific to a target subject, the models trained for other subjects were used as the prior models and were refined efficiently using Bayesian inference with a limited amount of data from the target subject. To train encoding models for a population, the models were progressively trained and updated with incremental data from different subjects. For the proof of principle, we applied these methods to functional magnetic resonance imaging (fMRI) data from three subjects watching tens of hours of naturalistic videos, while a deep residual neural network driven by image recognition was used to model visual cortical processing. Results demonstrate that the methods developed herein provide an efficient and effective strategy to establish both subject-specific and population-wide predictive models of cortical representations of high-dimensional and hierarchical visual features.

Deep recurrent neural network reveals a hierarchy of process memory during dynamic natural vision

The human visual cortex extracts both spatial and temporal visual features to support perception and guide behavior. Deep convolutional neural networks (CNNs) provide a computational framework to model cortical representation and organization for spatial visual processing, but unable to explain how the brain processes temporal information. To overcome this limitation, we extended a CNN by adding recurrent connections to different layers of the CNN to allow spatial representations to be remembered and accumulated over time. The extended model, or the recurrent neural network (RNN), embodied a hierarchical and distributed model of process memory as an integral part of visual processing. Unlike the CNN, the RNN learned spatiotemporal features from videos to enable action recognition. The RNN better predicted cortical responses to natural movie stimuli than the CNN, at all visual areas, especially those along the dorsal stream. As a fully observable model of visual processing, the RNN also revealed a cortical hierarchy of temporal receptive window, dynamics of process memory, and spatiotemporal representations. These results support the hypothesis of process memory, and demonstrate the potential of using the RNN for in-depth computational understanding of dynamic natural vision.

Deep Residual Network Predicts Cortical Representation and Organization of Visual Features for Rapid Categorization

The brain represents visual objects with topographic cortical patterns. To address how distributed visual representations enable object categorization, we established predictive encoding models based on a deep residual network, and trained them to predict cortical responses to natural movies. Using this predictive model, we mapped human cortical representations to 64,000 visual objects from 80 categories with high throughput and accuracy. Such representations covered both the ventral and dorsal pathways, reflected multiple levels of object features, and preserved semantic relationships between categories. In the entire visual cortex, object representations were organized into three clusters of categories: biological objects, non-biological objects, and background scenes. In a finer scale specific to each cluster, object representations revealed sub-clusters for further categorization. Such hierarchical clustering of category representations was mostly contributed by cortical representations of object features from middle to high levels. In summary, this study demonstrates a useful computational strategy to characterize the cortical organization and representations of visual features for rapid categorization.

Bootstrapping fMRI Data: Dealing with Misspecification

The validity of inference based on the General Linear Model (GLM) for the analysis of functional magnetic resonance imaging (fMRI) time series has recently been questioned. Bootstrap procedures that partially avoid modeling assumptions may offer a welcome solution. We empirically compare two voxelwise GLM-based bootstrap approaches: a semi-parametric approach, relying solely on a model for the expected signal; and a fully parametric bootstrap approach, requiring an additional parameterization of the temporal structure. While the fully parametric approach assumes independent whitened residuals, the semi-parametric approach relies on independent blocks of residuals. The evaluation is based on inferential properties and the potential to reproduce important data characteristics. Different noise structures and data-generating mechanisms for the signal are simulated. When the model for the noise and expected signal is correct, we find that the fully parametric approach works well, with respect to both inference and reproduction of data characteristics. However, in the presence of misspecification, the fully parametric approach can be improved with additional blocking. The semi-parametric approach performs worse than the (fully) parametric approach with respect to inference but achieves comparable results as the parametric approach with additional blocking with respect to image reproducibility. We demonstrate that when the expected signal is incorrect GLM-based bootstrapping can overcome the poor performance of classical (non-bootstrap) parametric inference. We illustrate both approaches on a study exploring the neural representation of object representation in the visual pathway.

Evaluation of Second-Level Inference in fMRI Analysis

We investigate the impact of decisions in the second-level (i.e., over subjects) inferential process in functional magnetic resonance imaging on (1) the balance between false positives and false negatives and on (2) the data-analytical stability, both proxies for the reproducibility of results. Second-level analysis based on a mass univariate approach typically consists of 3 phases. First, one proceeds via a general linear model for a test image that consists of pooled information from different subjects. We evaluate models that take into account first-level (within-subjects) variability and models that do not take into account this variability. Second, one proceeds via inference based on parametrical assumptions or via permutation-based inference. Third, we evaluate 3 commonly used procedures to address the multiple testing problem: familywise error rate correction, False Discovery Rate (FDR) correction, and a two-step procedure with minimal cluster size. Based on a simulation study and real data we find that the two-step procedure with minimal cluster size results in most stable results, followed by the familywise error rate correction. The FDR results in most variable results, for both permutation-based inference and parametrical inference. Modeling the subject-specific variability yields a better balance between false positives and false negatives when using parametric inference.