Session B Neuroengineering 2

Track 1

Session B

Neuroengineering 2

32 BME SENIOR DESIGN PROJECTS

A Classifier for Predicting Depth of Anesthesia Using Multimodal Cortical Recordings

Team 1: Vian Ambar Agustono, Elisa Cordeiro Lopes Technical Advisors: John A. White, Daniel Carbonero, Jad Noueihed

Anesthetics are widely used in modern medicine to induce loss of consciousness and henceforth allow painless performance of medical procedures that would be otherwise unbearable. Therefore, anesthetics have proven useful for surgeries and invasive clinical procedures in all populations. Because anesthetic response is not uniform across patients, it’s crucial to have an objective measure of depth, other than behavioral response Empirical measurements of anesthetic depth are difficult to take and are sometimes unavailable altogether. Hence, we propose a machine learning classifier to discern the level of anesthesia in mice using local field potential (LFP) and two-photon calcium imaging recordings. The voltage recordings are automatically preprocessed to extract burst suppression activity, as well as filter and identify artifacts. Recurrence rate and burst suppression rate are features chosen to characterize the nonlinear nature of burst suppression events which occur only in the anesthetized state. When a subject is anesthetized, neuronal signals shift in power from high to low frequencies, which is quantified as a feature by spectral edge frequency. These features were used in two classification models: random forest and support vector machine. The former is robust, flexible and offers greater accuracy due to its collection of prediction trees, while the latter is versatile and matches with the brain because the brain is highly nonlinear. We cross validate both models, to demonstrate both classification models are effective at predicting the depth of anesthesia in mice.

A Cloud-Based Framework for Organizing and Analyzing fNIRS Datasets

Team 3: Christian Arthur, Jeonghoon Choi, Jiazhen Liu, Juncheng Zhang Technical Advisors: David Boas, Stephen Tucker

Functional near-infrared spectroscopy (fNIRS) is a fast, safe and non-invasive neuroimaging technique that uses interaction between light and matter to study the brain and neural activity. Currently, fNIRS researchers use various available databases to store and share the data. In this project, we propose a cloud-based user interface that helps users in organizing their data and sharing methods in a standardized format. This framework utilizes a proposed fNIRS-BIDS data structure based on the Brain Imaging Data Structure (BIDS). This project includes the design and development of a front-end user interface along with the back-end Python packages for handling shared near infrared spectroscopy format (SNIRF) file and fNIRS-BIDS folders to be used in the cloud-based framework. A mock-up is developed with Figma to illustrate the functionalities and the accessibility of the user interface. The mock-up and the package is passed to BU Software & Application Innovation Lab for prototyping. The Python packages were thoroughly reviewed and currently maintained by the software engineers in the BU Neurophotonics Center. This development process provides grounds for implementing the standardized folder structure and processing pipelines that direct users in organizing their fNIRS datasets. Furthermore, this provides the foundation for implementation such as incorporating data quality metrics for immediate data feedback for the users. The long-term goal is to establish a standard platform that facilitates data sharing and quality assurance for fNIRS users.

Data Acquisition 1

Data Acquisition 2 Cloud Interface Upload

Cloud-based Framework

SNIRF

fNIRS-BIDS Conversion

Download Display/Edit

Data Acquisition n

Investigating the Bioenergetic and Biophysical Effect of Ultrasound on Neural Mitochondrial Activity

Team 7: John Rim, Rockwell Tang Technical Advisors: Xue Han, Emma Bortz, Yangyang Wang

Ultrasound neuromodulation is an emerging technique in the field of neuroengineering which uses acoustic pressure waves to control neural activity. Ultrasound stimulation is unique in its ability to pass through the cranium and deep tissue noninvasively, showing promise in clinical applications as a therapy for neurodegenerative diseases. Despite studies implicating mitochondria in the neural response to ultrasound, the role of mitochondrial activity in ultrasound neuromodulation have not been fully explored. Therefore, there is a critical need to determine the bioenergetic effects and biophysical mechanisms of ultrasound on neuronal mitochondria. The implication of a mitochondrial transduction pathway in ultrasound neuromodulation would broaden understanding of both the neuron’s response to mechanical stimulation and neural activity energetics, further enabling its use as an investigative tool and expansion of its clinical potential. First, an in vitro confocal imaging protocol was developed to quantify bioenergetic activity using the mitochondrial inner membrane-targeted voltage dye TMRM. A specimen setup was designed to enable simultaneous imaging and ultrasound stimulation. Then, the dynamics of the voltage gradient driving ATP production under spontaneous and pulsed ultrasound conditions were compared to assess the effects of stimulation. Finally, the biophysical interaction between ultrasound and the mitochondrial membrane was modeled in silico via molecular dynamics simulations, and membrane voltage changes were derived from the simulated deformations to be validated against empirical data. These results evaluate an alternate pathway for ultrasound transduction via the mitochondria and the possibility for bioenergetic uses of ultrasound stimulation.

NinjaNIRS 2022 Backpack System

Team 8: Robert Bing, Benjamin Lissner, Juan Luis Ugarte Nunez Technical Advisors: David Boas, Walker J. O’Brien, Bernhard Zimmermann

Brain monitoring devices like fNIRS and EEG offer valid approaches to monitor the cognitive states and activity of the brain. Moreover, it’s been found that utilizing fNIRS and EEG in tandem creates a hybrid fiber system that attains greater success than each individually. However, use of this improved method has been limited by its lack of portability and inability of long-term continuous monitoring of brain activity during movement, perception, and social interaction in real time while in the real world. Therefore, we propose a multimodal portable brain monitoring system that incorporates fNIRS, EEG, and an Eye Tracking System into a lightweight backpack that can capture the data in real time. The case to house the fNIRS system was designed using the CAD software, SolidWorks. The purpose-built casing was created to accommodate a greater array of optodes and updates on the custom NinjaNIRS system. The backpack was designed such that it safely houses each system and is ergonomic towards the user. An interior composed such that it will secure systems from damage, while maintaining accessibility to components for easy troubleshooting and data collection. Upon finalization of the realized hybrid backpack system, it’s within expectations that we’ll have found and designed a prototype that best exemplifies desired traits for long duration real time clinical brain recordings. The device will open up a whole new field of experimentation and testing from which valuable data can be collected and analyzed.