Development of a Whole-Cortex, fNIRS System for Perioperative Neuroimaging of Postoperative Delirium

Introduction: Every year, 51.4 million patients risk developing Postoperative Delirium (POD) following surgery, imposing an annual $33-152 billion cost on our healthcare system. POD is a poorly understood condition characterized by an acute change in cognitive ability and an increased risk of developing neurological disorders. The risk of POD increases with age, with 72% of patients aged 70-79 and 92% of patients aged 80-89 identified with POD following a surgical procedure. Past studies have strongly correlated hyperoxic cerebral reperfusion and reduced functional connectivity of cortical regions during surgery to delirium. Thus, perioperative monitoring of blood oxygen levels across the whole cortex could shed light on the risk factors of POD, reducing the overall burden and improving patient outcomes. However, many conventional neuroimaging techniques, including MRI/fMRI and PET scans, are bulky, expensive, and require the immobilization of the patient for long periods. This leads to many logistical challenges in the operating room (OR) and causes patient discomfort and significant noise artifacts in measurements. Additionally, current portable commercial systems (EEG/fNIRS) suffer from limited regions of coverage, high costs, and significant susceptibility to motion or sweat artifacts. To address this issue, we are developing a portable, wireless, whole-cortex functional near-infrared spectroscopy (fNIRS) system capable of performing real-time neuromonitoring in perioperative settings.

Materials and

Methods: To image the cerebral cortex, we position LEDs operating at 740 nm and 850 nm and silicon photodiode light detectors in an array approximately 3 cm apart, allowing optimal light penetration and scattering through the brain cortical tissue. Light pulses from the LEDs are emitted at the surface of the scalp into target tissue, and backscattered photons are detected by the detector probes. The analog signals from the photodiodes are converted into relative changes in deoxygenated hemoglobin (HbR) and oxygenated hemoglobin (HbO) concentration using a modified Beer-Lambert equation, facilitating quantitative analysis of cerebral hemodynamics, which is used to infer neural activity. Custom-designed controller boards, each with a low-cost Arduino microcontroller unit and a sensor board equipped with 8 sources and 16 detectors, are used with the end goal of connecting multiple boards simultaneously for comprehensive whole-cortex coverage. These boards interface wirelessly using Bluetooth Low Energy (BLE) technology with EasyNIRS, an app developed in MATLAB App Designer, streamlining data collection, live visualization, and data transfer. Standard laptop hardware was used to run the app, making it feasible to scale the system for whole-cortex applications, managing increased data load and processing demands for real-time functionality and extensive neuroimaging studies.

Results, Conclusions, and Discussions: To validate the electronics of the device, a linear relationship between LED driving power and current was ensured, with coefficients of variation maintained below 0.08. Additionally, it was confirmed that the LEDs could effectively pulse at 4 millisecond intervals to produce characteristic square waveforms, effectively eliminating crosstalk between different light sources during data collection. Next, the photodiodes were verified to operate within their linear range under low light conditions, ensuring accurate signal detection. Preliminary breath-holding and arithmetic test data collected using a 33-channel prototype demonstrated the ability to reproduce well-characterized cerebral hemodynamic plots upon the introduction of stimuli. This will be relevant as we move on to more rigorous validation tests, such as finger-tapping tests and auditory tests. Next, a pressure-standardized, tissue-mimicking phantom apparatus was employed to develop a platform for cross-validation and consistent performance analysis of the electronics. The phantom apparatus was shown to characterize the effects of varying contact pressure on signal strength within a 4-8% error margin at constant pressure values using an epoxy-based tissue phantom. These results collectively validate the prototype device, and future work will focus on enhancing the system’s data rate, currently at 7 Hz, as well as enabling BLE connections to multiple controller boards simultaneously to achieve whole-cortex coverage. Such a system would have the potential to identify real-time clinical events as risk factors for POD and reduce the response time between diagnosis and treatment, directly addressing the limitations of conventional neuroimaging techniques.

Acknowledgements (Optional): This work was conducted at the Bowden Biomedical Optics Laboratory under the guidance and instruction of Professor Audrey Bowden and graduate student Seth Crawford.