Reliability and similarity of resting state functional connectivity networks imaged using wearable, high-density diffuse optical tomography in the home setting.

BACKGROUND: When characterizing the brain's resting state functional connectivity (RSFC) networks, demonstrating networks' similarity across sessions and reliability across different scan durations is essential for validating results and possibly minimizing the scanning time needed to obtain stable measures of RSFC. Recent advances in optical functional neuroimaging technologies have resulted in fully wearable devices that may serve as a complimentary tool to functional magnetic resonance imaging (fMRI) and allow for investigations of RSFC networks repeatedly and easily in non-traditional scanning environments. METHODS: Resting-state cortical hemodynamic activity was repeatedly measured in a single individual in the home environment during COVID-19 lockdown conditions using the first ever application of a 24-module (72 sources, 96 detectors) wearable high-density diffuse optical tomography (HD-DOT) system. Twelve-minute recordings of resting-state data were acquired over the pre-frontal and occipital regions in fourteen experimental sessions over three weeks. As an initial validation of the data, spatial independent component analysis was used to identify RSFC networks. Reliability and similarity scores were computed using metrics adapted from the fMRI literature. RESULTS: We observed RSFC networks over visual regions (visual peripheral, visual central networks) and higher-order association regions (control, salience and default mode network), consistent with previous fMRI literature. High similarity was observed across testing sessions and across chromophores (oxygenated and deoxygenated haemoglobin, HbO and HbR) for all functional networks, and for each network considered separately. Stable reliability values (described here as a <10% change between time windows) were obtained for HbO and HbR with differences in required scanning time observed on a network-by-network basis. DISCUSSION: Using RSFC data from a highly sampled individual, the present work demonstrates that wearable HD-DOT can be used to obtain RSFC measurements with high similarity across imaging sessions and reliability across recording durations in the home environment. Wearable HD-DOT may serve as a complimentary tool to fMRI for studying RSFC networks outside of the traditional scanning environment and in vulnerable populations for whom fMRI is not feasible.

Transfer Learning from Healthy to Unhealthy Patients for the Automated Classification of Functional Brain Networks in fMRI

Functional Magnetic Resonance Imaging (fMRI) is an essential tool for the pre-surgical planning of brain tumor removal, which allows the identification of functional brain networks to preserve the patient’s neurological functions. One fMRI technique used to identify the functional brain network is the resting-state-fMRI (rs-fMRI). This technique is not routinely available because of the necessity to have an expert reviewer who can manually identify each functional network. The lack of sufficient unhealthy data has so far hindered a data-driven approach based on machine learning tools for full automation of this clinical task. In this article, we investigate the possibility of such an approach via the transfer learning method from healthy control data to unhealthy patient data to boost the detection of functional brain networks in rs-fMRI data. The end-to-end deep learning model implemented in this article distinguishes seven principal functional brain networks using fMRI images. The best performance of a 75% correct recognition rate is obtained from the proposed deep learning architecture, which shows its superiority over other machine learning algorithms that were equally tested for this classification task. Based on this best reference model, we demonstrate the possibility of boosting the results of our algorithm with transfer learning from healthy patients to unhealthy patients. This application of the transfer learning technique opens interesting possibilities because healthy control subjects can be easily enrolled for fMRI data acquisition since it is non-invasive. Consequently, this process helps to compensate for the usual small cohort of unhealthy patient data. This transfer learning approach could be extended to other medical imaging modalities and pathology.