Derivative temporal clustering analysis: detecting prolonged neuronal activity

Abstract

Temporal clustering analysis (TCA) and independent component analysis (ICA) are promising data-driven techniques in functional magnetic resonance imaging (fMRI) experiments to obtain brain activation maps in conditions with unknown temporal information regarding the neuronal activity. Although comparable to ICA in detecting transient neuronal activities, TCA fails to detect prolonged plateau brain activations. To eliminate this pitfall, a novel derivative TCA (DTCA) method was introduced and its algorithms with different subtraction intervals were tested on simulated data with a pattern of prolonged plateau brain activation. It was found that the best performance of DTCA method in generating functional maps could be obtained if the subtraction interval is equal to or larger than the length of the rising time of the fMRI response. The DTCA method and its theoretical predication were further investigated and validated using in vivo fMRI data sets. By removing the limitations in the previous TCA, DTCA has shown its powerful capability in detecting prolonged plateau neuronal activities.

Introduction

In a typical functional magnetic resonance imaging (fMRI) study, assumptions about temporal activation patterns can be made through an explicit experimental paradigm. For example, it is expected that the visual cortex will be activated when visual stimuli are presented. Paradigm-dependent data processing methods can be used in such situations to localize the brain activation. Group t test and cross-correlation are commonly used paradigm-dependent data processing methods.

However, in situations where little or no temporal information is available for making inferences about neuronal activity, conventional paradigm-dependent data processing methods are significantly limited. These circumstances are most likely found in fMRI studies involving brain activation induced by drugs, nutrition, sleep or epileptic seizure. In these conditions, paradigm-independent (data-driven) processing methods have been developed. Paradigm-independent methods are able to produce functional maps without prior knowledge of the temporal characteristics of neuronal activity. Current paradigm-independent methods used in neuroimaging include principle component analysis [1], [2], [3], independent component analysis (ICA) [4], [5], fuzzy clustering analysis [6] and temporal clustering analysis (TCA) [7], [8], [9]. Previously, we demonstrated that TCA is comparable to the most commonly used ICA techniques in generating functional brain maps in event-related fMRI experiments [10]. The TCA method has been employed in studies of eating [7] and epileptic activity [11]. The TCA method is gaining popularity due to intrinsic advantages in its computational efficiency given the simplicity of data interpretation, particularly when compared to the uncertainty of selecting a true activation pattern from numerous components generated in ICA analyses. However, despite its capability in generating functional maps for transient brain activation, the original TCA method is relatively insensitive to prolonged plateau activation patterns [8]. This pitfall results from the fundamental basis of the TCA algorithm. TCA detects brain activity by arrogating the number of voxels with the maximal extremities at each time point. Thus, transient neuronal activity produces a sizable peak in the TCA curve and is readily detectable. However, prolonged brain activation distributes the maximal extremities across the activation period and is poorly specified in TCA output. Given that prolonged brain activation patterns are commonly observed in pharmacologic investigations (e.g., heroin or cocaine challenges increase brain activity on the order of minutes [12], [13], [14]) and during the rapid eye movement phase of sleep [15], we introduce a variant to the TCA method, derivative TCA (DTCA), designed specifically to detect prolonged brain activation. This DTCA method exploits the observation that the onset of prolonged plateau neuronal activity will produce a detectable peak in the first derivative of an fMRI time series. Here, we present the DTCA theory and test the method with computer-simulated and in vivo fMRI experimental data sets.