Using MazeSuite and Functional Near Infrared Spectroscopy to Study Learning in Spatial Navigation

MazeSuite is a complete toolset to prepare, present and analyze navigational and spatial experiments. Functional near-infrared spectroscopy (fNIR) is an optical brain imaging technique that enables noninvasive and portable monitoring of cerebral blood oxygenation changes. This paper summarizes collective use of MazeSuite and fNIR within a cognitive processing learning paradigm.

Our goal In this presentation is twofold. First, to illustrate the experimental protocol design process and the use of May suite and second, to demonstrate the setup and deployment of the FNIR Brain Activity Monitoring System. May suite can be used to design and edit adapted 3D environments, as well as track A participant's behavioral performance.

To demonstrate these, a sub-sample from a study is reported to show how to use both May suite and FNIR in a single experiment. The study involves the assessment of cognitive activity of the dorsal lateral prefrontal cortex during the acquisition and learning of computer mace tasks for blocked and random practice Orders. FNIR Is an optical brain monitoring technique that uses near infrared, like to track changes in hemodynamic responses within the cortex.

Non-invasive monitoring of cerebral hemodynamics and oxygenation were first demonstrated by job ossy and colleagues in 1977. This technology allows the design of portable, safe, affordable, non-invasive, and minimally intrusive monitoring systems that can be used to measure brain activity in natural environments. Ambulatory and field conditions near infrared light can penetrate tissue a few centimeters within 700 to 900 nanometers.

When light enters the tissue, it is either absorbed or scattered. Absorption is due to chromo force light absorbing molecules, such as hemoglobin and water scattering occurs in the structure of the tissue such as cell membranes and layers. A typical FNIR measurement contains a measurement unit that has light sources and detectors.

When a light source activates light is introduced over the scalp, penetrates it and travels in all directions in the tissue and loses intensity as it travels, some of the light reaches back to the surface because of multiple sequential scattering and it is registered by the detector, the photons that reach the detector actually travel through a banana shaped volume. Measurements at the detector provide information about this volume between light source and detector. Couple called an oid, the depth of the penetration, in other words, the curvature of the banana is a function of the distance between the light source and detector.

That's why FNIR sensor geometry is key in the design. Depending on the type of measurements for cognitive task and measurement from the prefrontal cortex, opto separation is usually 2.5 centimeters to three centimeters. In this study, we have used FNIR devices model 1000.

That is based on the designs of chance and colleagues in the 1990s and further developed at the Optical Brain Imaging Lab of Drexel University. This instrument does not rely on fibers or light guides to interface optos with the skin. Hence, it is easier to set up, more comfortable for long sessions and less prone to movement artifacts.

However, this sensor is designed specifically for detection of cortical activation of the dorsal lateral prefrontal cortex that is under the forehead and cannot be used in other head regions because of interference from hair. The FNIR sensor pad used in the study contains four light emitting diodes that shine non-coherent light at 730 nanometers and 850 nanometers. There are 10 photo detectors and by shining LEDs in a sequential order, along with using surrounding detectors 16 measurement locations.

Boxes are being sampled at each scan. This includes light intensity measurement of two different wavelengths and also a dark measurement for ambient light, totaling three channels for each measurement location, so in total, there are 48 channels recorded from the scan. The sensor pad is placed over the forehead of the subject side.

One should be on the subject's left and side two. On the subject's right side, the sensor should be placed just over the eyebrows and should be centered vertically. The imaginary vertical symmetry line passes through the midline of the forehead and then the nose.

The sensor pads center line coincides with the midline of the forehead and nose. Once the sensor pad is positioned, the cables are pulled on the two sides and connected at the back of the head. With the cable secured with the clip, it is critical to check that the sensor pad is properly coupled with the skin and that there are no bumps or spaces between the optos and skin.

A good way to check for proper coupling is to apply a little pressure over the sensor pad and feel if positioning changes with pressure. Sometimes a headband, elastic, and or firm cloth, such as a tennis band or bandana, can be placed over the FNIR sensor to secure coupling of the optos with the skin. After the sensor's position, signals at all channels should be verified by starting data acquisition First, the sensor pad should be connected to the FNIR hardware control box.

The device should be connected to the computer by USB cable, and both systems should be powered on. Next, run Kobe Studio on the computer by clicking on the shortcut at the left pane, click on the start current device link. If the settings are correct, the message pane will display that the device had been started and graphs will display the newly acquired signals.

The signal levels depend on the LED current and gain settings. A good rule of thumb is to have these parameters above 700 millivolts and below 4, 000 millivolts. In addition, signals should be stable.

Too much variation and spikes may indicate improper sensor coupling or cable or hardware connectivity problems. In some cases, one or two lateral channels, one and two on the left and 15 and 16 on the right might be placed over the hair and their signal values are too low. You may just continue and eliminate those channels later in the analysis.

You may need to adjust settings based on each subject since there may be high individual differences due to different optical properties of an individual skin. To change device settings first, click on stop current device. Then click on device properties in the left pane.

Go to the data acquisition settings tab in the dialogue box. If you would like to increase signal values first, increase the LED current value. An increase in the LED current value means that the LEDs will shine brighter.

If you would like to degree signal levels first, decrease the gain. Initial gain value is used for all voxel. After you set the values, click save and then click start.

Current device to start data acquisition with a newly set parameters, it is also common and useful to use markers to identify certain events during experiments. There are two types of markers, manual and automatic. Manual markers are generated by button clicks on the main window of K, and they are time tagged and saved with FNIR data.

Automatic markers are received from an external device or computer software to receive automatic markers while the device is stopped. Go to device settings located on the left pane and within the synchronization settings tab check, listen for markers. Serial port is a suggested method for marker communication.

Verify that the serial port number is set to a valid port number on the computer. Then stimulus software such as ePrime or May Suite can be sent to send markers by values that are received time tagged and saved along with FNIR data on this computer. Raw FNIR Signals are time series signals that are light intensity.

Noise in the data can be removed before or after conversion of raw light.Intensity. Values to oxygenation values the physiologically irrelevant data such as respiration, heart pulsation, and equipment noise need to be eliminated from the raw FNIR measurements. The heart rate related component usually has a peak that is around or above 0.5 hertz and respiration is between 0.2 hertz to 0.4 hertz to eliminate these physiological artifacts.

Finite impulse response and linear phase low pass filters are used. Signals might also be corrupted by motion artifacts. When the FNIR sensor light sources and or detectors slide from their original attached location or lose contact with the skin due to head motion, unexpected sudden burst or spikes can happen in the FNIR measurements.

Furthermore, if the light source loses coupling with the skin, the detector may record either very low values since no light may pass towards it or extreme high intensities and momentary saturation due to the reflected light from the surface of the skin. Similar saturation effects may happen if the detector pops and loses contact with the skin. Causing ambient light to leak head movement can further cause changes in the pressure applied to the sensor pad or to the light sources and detectors.

These changes may allow more photons to enter into the tissue, thus temporarily varying the detected light intensity. Other than just visual inspecting the data for possible motion artifacts, there are a growing number of motion artifact detection and removal algorithms to automate the process and eliminate subjectivity Artifact removal algorithms in the literature ranges from simple lowpass and VAM pass filters to wavelet analysis from independent or principle component analysis to optimal filtering such as adaptive wiener and cowman filters Continuous wave systems. Raw FNIR signals are converted to the relative oxygenation changes by using the modified beer Lambert Law.

Optical density at a specific input wavelength is the logarithmic ratio of input light intensity and output light intensity OD is also related to the concentration and absorption coefficient of chromo force, the corrected distance of the light source and detector plus a constant attenuation factor having the same input light intensity at two different time instances. The difference in OD can be written in terms of only detected light intensity values. Typically, the two wavelengths chosen are one within 700 to 900 nanometers where the absorption of oxyhemoglobin and deoxyhemoglobin are dominant as compared to other tissue chromophore and two one below and one above the iso spastic, which is approximately 805 nanometers where absorption spectrums of deoxy and oxyhemoglobin cross each other.

This equation can be set to solve for concentrations if the two by two matrix is non-significant. After relative changes of oxygenated and deoxygenated hemoglobin values are calculated. The next step is to extract features depending on the experimental protocol and cognitive task used.

Feature extraction is most commonly used to reduce the amount of data and to make comparisons between different cognitive tasks, subject groups, and anatomical locations using statistical analysis. Commonly used features involve maximum minimum mean or medium value of the oxygenated and deoxygenated hemoglobin and reaction time, which is the time elapsed until the minimum or maximum is reached. These features can be extracted within an evoked hemodynamic response to certain cognitive stimulus obtained through the use of single trial paradigms or with data epoch or block of data corresponding to certain conditions implemented through block trial designs and paradigms.

May suite Consists of three components each targeted towards a specific stage of the experiment. Design, experimentation, and analysis. Maze Maker allows the user to easily design a 3D environment for each stage of the experimental protocol.

Simple environments can be created in moments. First, define the floor area and then use the wall tool to define the boundaries of the maze. Finally, place a subject starting position and then exit area.

After saving, the quick run tool can be used to immediately test the environment. More complicated environments can be created to meet the requirements of a particular experimental design. Environments can be customized with the introduction of interactive objects, lighting controls, and texturing.

Using the Maze List Builder Mel files can be created and saved that combine particular combinations of mazes and user messages for use as experimental procedures. Masis functionality can be used to carry out the specific experimental design levels can be arranged with increasing or randomized difficulty as a storyboard or with control versus experimental levels. Individual trials of the experiment are run through the use of the Maze Walker program.

Changing settings in Maze Walker allows users to further specify the conditions of the experiment. The control can be changed to accept input from different devices including keyboard, mouse, and joystick. External devices can engage in two-way communication with Maze Walker gathering information or triggering changes within the environment.

Device communication can be handled over T-C-P-I-P or serial cable and can interface with a wide variety of devices including E-E-G-F-M-R-I and FNIR. During sessions, mais Walker would log the subject movement along with any events that occurred during the session. With the Maze Analyzer program, MACE files can be displayed with a recorded user path for review.

Different paths can be overlaid to help in analysis and statistical information about different trials can be calculated. Additionally, collected information can be exported to Excel or MATLAB for additional processing. Our cognitive Repertoire includes a wide array of functions and abilities that can be accessed through the use of a variety of tasks, including virtual spatial navigation, mazes brain activation Patterns in the frontal cortex varies from the initial novel task performance after practice, and during retention.

Using FNIR and focusing on the prefrontal cortex, we are capitalizing on the function of the PFC, especially executive function, the regulation of cognitive function in action, the use of PFC during problem solving tasks and the use of higher level structures while maintaining multiple items in working memory. The organization of practice when learning multiple similar tasks is a learning phenomenon called the contextual interference Effect. Effects of the contextual interference are evident when individuals acquire multiple tasks on different practice schedules.

High contextual interference is created when tasks to be learned are presented in a non-sequential unpredictable order. Low contextual interference block practice is created when the tasks to be learned are presented in a predictable order. Each performed 315 acquisition trials, 105 trials of each of the three mazes across three days, Monday, Wednesday, and Friday.

The following Monday, 72 hours following acquisition, 30 retention and 20 transfer trials were performed in a random order. PFC activity was monitored during all phases for 16 opto sites. Using FNIR dependent measures included relative changes in the mean oxygenated, hemoglobin, and behavioral measures of total time, path, length, and average velocity for the mazes.

The behavior results show that for both practice orders, there's a monotonic decreasing trend for maze time, the total time necessary to complete the maze suggesting as the participants practice, they completed the maze in shorter periods of time, which is expected for behavioral assessments of learning. In addition, the maze velocity, the average speed at which participants navigate the maze increases with practice. Again, an improvement in behavioral measures across time is expected as inferences of learning for these results.

There is side-by-side comparison of the random versus block practice for each transfer task for the easier transfer task. May four block practice outperformed random practice. However, for the more difficult transfer task, maze five, random practice was superior to block mean oxygenated.

Hemoglobin concentration changes during practice trials indicate that the blocked order required higher brain activation compared to the random order. In addition, when comparing within a practice order for the blocked order learning, a new task, as in the transfer phase required higher brain activation. Given that the practice order was different for the tasks already learned.

For the participant that learned the task in a blocked order, this stratified random practice order may have been sufficiently novel to require additional effort and cognitive resources to perform the task. Moreover, for the random practice order, the transfer phase of neural activation was not higher than the retention phase. In Conclusion, this exploratory study demonstrated the use of May suite and FNIR to study neurobehavioral aspects of learning in spatial navigation.

May Suite enables the design and application of simple 3D environments with a user-friendly graphical interface, and it automatically records behavioral measures for within subject or across subject comparisons. FNIR is a portable, safe, and non-invasive brain monitoring tool that has been used in clinical laboratory and natural settings to study brain activation. We hope this presentation was helpful in explaining these tools.

Good luck with your experiments.