Non-Invasive Multi-Modal Imaging to Evaluate Disease Severity in Sickle Cell Disease

Background: Pain is one of the major symptoms of sickle cell disease (SCD). Sickle pain can include both nociceptive and neuropathic components and can start from infancy. Unfortunately, pain management remains challenging in these patients partly due to the lack of understanding and tools to non-invasively map brain function. Pain can alter neural networks and in turn neuropathic changes can augment pain leading to chronic pain. To analyze brain function we have developed a non-invasive electroencephalography (EEG) coupled to functional magnetic resonance imaging (fMRI) to map how chronic pain affects neurological changes in the brain of SCD patients.

Methods: Simultaneously recorded EEG and fMRI during resting state were recorded in patients with SCD (N = 11) and healthy controls (N = 13). MR-compatible amplifiers and a 64-channel EEG cap were used to record the EEG data inside a 3T Siemens Trio MR scanner. The fMRI data were recorded with a 16 channel head coil and an echo-planar imaging sequence. The fMRI data were preprocessed using SPM8 software, while EEGLAB was used to preprocess the EEG data. A group independent component analysis (ICA) was performed on the fMRI data using GIFT software. The ICA revealed 9 resting state networks (RSN) observed in both patients and controls. A spectrogram was used on the EEG data to obtain power time courses for different frequency bands (including delta, theta, alpha, beta1, and beta2). EEG-fMRI analysis was performed using the power time courses and comparing them across all voxel time courses using a general linear model. Group level results were generated from single subject maps using height and extent thresholds of p < 0.001. Any regions of interest (ROI) for the EEG-fMRI power analysis were identified using contrast image maps between the control and patient groups. The z-score of the ROI was obtained from back-projected RSN maps. Z-scores were then correlated to clinical data to identify any relation between neurological measures and disease severity.

Results: The RSN identified by the group ICA included the default mode network, salience network, sensory motor network, and others. The EEG-fMRI power analysis showed that patients have greater activation in the insula and rolandic operculum compared to controls in the beta1 frequency band. This result was confirmed when a contrast image of "patients > controls" was obtained and a ROI was identified in the left insula. The primary peak location of this ROI was (x = -36, y = 2, z = 13) and the z-score of the primary peak was 2.99. Due to the overall layout of the group activation map for the beta1 frequency and the location of the ROI, the salience network was selected to be the RSN studied. Back-projected salience network maps were used to obtain individual z-scores from the left insula ROI from controls and patients. The individual z-scores were compared to clinical data and a significant correlation was found between the mean z-score and age for both patients (p < 0.005) and controls (p < 0.005). However, while controls showed a strong negative correlation between z-score and age, patients showed a positive correlation between z-score and age.

Conclusions: These results indicate that the beta1 frequency activation observed in patients is most likely due to the salience network, which is theorized to be responsible for processing external input, including pain. We found that patients have different neurological activation compared to controls within common EEG bands. Furthermore, the left insula ROI z-scores increased with age in patients, indicating that the left insula's role in the salience network may be stronger as patients age. This may be due to disease severity, and hence pain, increasing with age in SCD. The opposite trend was observed in controls, where the role of the left insula in the salience network seems to decrease in strength with age. This most likely reflects altered connectivity of RSN due to normal aging. Our results suggest that altered behavior in beta1 can be used as a biomarker of disease severity, and that non-invasive imaging techniques can be used to identify biomarkers of disease severity. This work was supported in part by NIH grant U01-HL117664 and NSF IGERT grant DGE-1069104.