Theses supervised by Prof. Dr. Burak Güçlü
10 theses · Boğaziçi University
Sıçanlarda yürüme parametreleri ve makine öğrenmesi ile omurilik yaralanmasının sınıflandırılması
Spinal cord injury (SCI) represents a critical neurological condition with high morbidity, significantly impacting sensory motor functions. This thesis introduces a novel approach to classify the time after SCI and rats which recieved neuromodulation therapy by using machine learning (ML) based analysis of gait parameters and locomotor scores. Utilizing data derived from video imaging of rat locomotion, this study evaluates the accuracy and feasibility of the approach compared to previous literature. Key parameters analyzed include stance duration, swing duration, stride distance, limb duty factor, and paw area, alongside traditional metrics like Basso, Beattie and Bresnahan (BBB) locomotor rating scale and Von Frey (vF) withdrawal thresholds. The integration of markerless pose estimation tool, DeepLabCut (DLC), allowed for detailed extraction of gait parameters, overcoming challenges associated with lateral plane-focused methodologies. By using images from lateral and bottom views, footsteps were associated with animals' body movement and the gait pattern was extracted by deep-learning methods. Gait parameters, BBB scores and vF thresholds were input as features to Ensemble Learning including various ML methods to predict time after injury and neuromodulation treatment. In animals with SCI, gait parameters and BBB scores from both fore-hindlimbs, allowed prediction of post injury class (6 end-points) with \%43 accuracy. The same feature set predicted treatment with \%71 accuracy. Results showed that the methodology can effectively differentiate between treatment groups and time after SCI. However, additional work is needed to improve accuracy. Parameter extraction by deep-learning also provides an accessible and cost-effective solution for SCI research.
Omurilik yaralanması olan sıçanlarda periferik transkütanöz elektriksel sinir stimülasyonu ve yürüme analizi
Electrical stimulation strategies are investigated in the literature to recover sensorimotor impairments in spinal cord injury. In this study, transcutaneous electrical nerve stimulation (TENS) (pulse width: 0.3 ms, pulse frequency: 2 Hz, amplitude: 2 × motor threshold, duration: 30 min. at 3 days/week) was applied bilaterally to the tibial nerve rats with spinal cord injury (SCI). The study included sham (n=10), SCI (n=15), NS-sham (n=6), and NS-SCI (n=18) with endpoints at day 1 (D1), day 7 (D7), month 1 (M1), month 2 (M2). After T8-T9 laminectomy, contusion-type SCI was induced by using a computer-controlled custom-made impactor device (contactor diameter: 2.3 mm, peak force: 0.9-2 N, displacement: 1.75 mm, duration: 0.5 s). The severity (moderate-to-severe) of the injury and recovery was measured Basso, Beattie ve Bresnahan (BBB) locomotor rating scale. The mechanical withdrawal threshold was measured by von Frey filaments. DeepLabCut system was used to track each paw from high-speed (120 fps) video recordings and gait parameters (stance/swing/stride durations, limb duty factor, stride length, footprint area) were extracted in Matlab before detrending for speed. Average BBB scores were significantly decreased after the injury (D1: 0). The recovery was higher at NS-SCI (BBB: 17.5) than SCI (BBB: 8) (p = 0.003) at M2. vF scores also improved but were not statistically different between SCI and NS-SCI (p = 0.495). In gait parameters, the facilitative effect of TENS was observed up until D14; however, there was a reversal trend in limb duty factor and footprint area after D14. Overall, TENS had positive effects on locomotor function and stepping after SCI. However, synchronized application of TENS with stepping may have better results in terms of gait parameters in future studies.
Functional characterization ofgraphene-based thin-film microelectrodeson rat sensorimotor cortex
Neuroprostheses based on cortical implants are promising to provide partial sensorimotor function in severe neurological conditions such as spinal cord injuries and amyotrophic lateral sclerosis. One of the key components of these systems is the microelectrode array, which is used for recording brain activity to control a robotic limb and/or for stimulation to induce somatosensory feedback. Graphene is a good candidate as electrode material due to its intrinsic features such as high electrical conductivity and charge injection capacity, high mechanical strength, flexibility and biocompatibility. Evoked local field potentials were recorded epidurally at the hindpaw representation of SI in anesthetized Wistar albino rats. The vibrotactile stimuli were bursts of sinusoidal (5-, 40-, and 250-Hz) displacements (duration: 0.5-s, amplitude range: 19 - 270 $\mu$m) applied on the glabrous skin. Performance comparisons were made between matching research grade graphene and commercial Pt-Ir surface electrodes on the same subjects (active site diameter: 25-$\mu$m). Robust evoked potentials could be observed shortly after the onset of contralateral stimuli in both electrodes. Pt-Ir electrodes exhibited slightly higher SNR while the lowest impedances were recorded from the channels of the graphene array. Variance of the impedance values were smaller for the channels of the Pt-Ir electrodes. The performance of the graphene electrode channels was observed to be heterogeneous due to ongoing development efforts. This thesis includes one of the first functional tests of graphene electrodes during processing of the natural sensory stimuli in the brain.
Design of a virtual environment for discrimination tasks to determine the psychometric function
In this thesis, we used the Gazebo simulation software to develop a virtual environment for interactions of objects to perform psychophysical tasks to distinguish between objects at differing stiffnesses. The virtual environment is composed of a probing device, a cube, and a slider. The probing device is used to touch the cube which moves on the slider. The environment includes objects of different stiffnesses, and the user attempts to discriminate between these objects. To simulate the effect of deformation in the cube, we modeled the cube as a mass-spring-damper system, so the object generates a force proportional to the distance it is pushed by force generated from the probe contacting the cube. The virtual environment was tested by recording the step response of each cube with different k constants. The cube deformed according to the model. Six users were asked to perform psychophysical trials on the virtual environment for stiffness discrimination. Six identical cubes with different stiffnesses were used, where the stiffest cube was used as the reference. The subjects were asked to discriminate between each cube and the reference object. Then we plotted the psychometric curve and determined the discrimination threshold from the data produced. The experiment was done in three phases with twenty trials in each phase. In the first phase, the user would see the virtual environment and also listen to the auditory signal. In the second phase, the user was blindfolded and only received the audio signal. In the third phase, the auditory signal of the virtual environment was muted, and the user only received the visual signal, which was the distance the cube moved. The psychophysical trials show that the virtual environment can be used to determine the psychometric curve and discrimination threshold of the user's ability to discriminate between objects of different stiffnesses. Keywords: Psychophysics, Psychometric Function, Virtual Environment, Stimulus, Discrimination Threshold.
Tactile processing and vibrotactile discrimination capacity in children with tourette syndrome
Tourette Syndrome (TS) is a childhood-onset developmental psychiatric disor- der. Pediatric patients are diagnosed with TS if they show multiple motor tics and at least one vocal tic for at least one year. The tic severity is known to be reduced in most of the cases as the patient progress into adulthood which suggests a cerebral adapta- tion over time. The pathology of TS is not clear; however, neurotransmission deficits, especially of γ-aminobutyric acid (GABA), and structural alterations in the cerebral structures are believed to be play a role in disorder's occurrence. Existing literature suggests the tics to arise from hyperexcitability due to GABAergic dysfunction, and the adaptive somatosensory mechanisms in TS to be disrupted. This study aimed to extend the GABAergic adaptive dysfunction in TS hypothesis by assessing the detec- tion and difference thresholds through a psychophysical vibrotactile battery. Thirty TS children (7 female, 23 male) and 25 healthy controls (7 female, 18 male) participated in the experiments. Vibrotactile stimuli were generated by a portable device and applied to the fingertips of the subjects. The vibrotactile battery consisted of Choice Reaction Time (cRT, amplitude: 200 μm, Static Detection Threshold (DT_s ), Dynamic Detec- tion Threshold (DT_c , amplitude ramp: 2 μm/s ), Amplitude Discrimination (AD, standard stimuli: 50, 100, 200 μm ), and Amplitude Discrimination with single-site adaptation (cAD, the same standards, adapting stimuli: 100, 300 μm, adaptation dura- tion: 1 s) measurements. The analyses showed that both groups produced comparable detection thresholds. Amplitude discrimination tasks produced further support for the GABAergic adaptive dysfunction in TS hypothesis, since in the baseline AD tasks TS group produced significantly higher difference thresholds, and in the cAD tasks control group closed the gap by showing a more prominent adaptation.
Prediction of psychophysical responses from spike recordings in rat sensorimotor cortex by using Bayesian models
In this thesis, we studied the fundamental question in neuroscience: how perception is built based on the sensory stimuli from the physical world and turned into motor actions in the face of uncertain neural representations. The vast body of literature contains models using neural activity to decode stimulus parameters, motor responses, and behavioral patterns. In particular, this line of research became more important as sensorimotor neuroprostheses and brain-computer interfaces (BCI) were made possible by recent advances in technology. The real-time algorithms used in those applications have many limitations. The main goal of the thesis is to use Bayesian models to understand sensorimotor processing and develop a novel approach for future BCIs. Specifically, spike data were collected from awake behaving rats during psychophysical yes/no detection task. Within a Bayesian framework, task-related priors, posterior beliefs, and the objective function to match the observed choice of the animal were calculated. The random variables for stimulus presentation, population neural activity, and motor responses were combined in a probabilistic graph network. First, a somatosensory neuroprosthesis application is demonstrated. Next, the Bayesian model was used to predict trial-by-trial responses offline. It was found that psychophysically low-performing rats could be modelled better with the Bayesian approach. The simulation results were compared to predictions of other supervised learning algorithms (such as linear discriminant analysis, decision trees, etc.). The Bayesian prediction was one of best among those algorithms for low-performing rats. Finally, behavioral responses from previous trials and neural activity from the current trial were included in various Bayesian models, which studied the effects of incremental information to predict the behavioral response in the current trial. The results showed that the average firing rates in a population of neurons are mostly adequate to predict lever presses in the psychophysical task with high sensitivity and low bias. This thesis provides new insights into computational modeling to understand sensorimotor processing and development of future BCIs. Bayesian modeling can be particularly useful in rehabilitation and during the training period of neuroprostheses.
Effects of prior stimulation on tactile evoked epidural field potentials in rat S1 cortex
Understanding how tactile sensation is processed in the somatosensory cortex is crucial for the development of neuroprostheses that can provide a realistic sense of touch. Exploring the electrophysiological basis of vibrotactile forward masking offers valuable insights into how the brain integrates and responds to sequential sensory inputs. This understanding can help drive progress in the development of haptic interfaces and the enhancement of neuroprosthetic technologies designed to improve tactile perception. Epidural field potentials were recorded from the hind paw representation of the rat S1 cortex by using various experimental parameters. The effects of the prior stimulus on the test stimulus window measured as dB difference and latency difference were evaluated. The results indicated that all main factors had a significant impact on the dB difference. An increase in the amplitude of the prior stimulus was found to enhance suppression effects. The suppression decreased as the temporal gap increased. By demonstrating the impact of the prior stimulus, the study underscores the fundamental influence of preceding sensory inputs in shaping subsequent sensory processing.
Design of a vibrotactile balance support system with a virtual reality training program
This thesis aims to design a vibrotactile feedback(VTF) system to help balance rehabilitation with virtual reality (VR) training. First, the training program was built in a virtual reality platform by using Unity3D and Blender software. Data for visual and vibrotactile psychophysical limits were obtained in several experiments performed by one participant. The VR platform simulated anterior/posterior sways which were conveyed to the participant visually and/or by VTF. Visual experiments consisted of motion detection, angle discrimination, and angular velocity discrimination of an avatar in the VR screen. All motion detection thresholds were found to be lower than 0.04 deg/s. Angle and velocity discrimination limens were in the range of 0.26-0.46 deg and 0.19-0.34 deg/s in the visual avatar. Arduino UNO was used to control six vibration motors placed around the upper arm . Motors were recruited incrementally as the avatar's sway angle increased. Angular velocity was mapped either by mixed (Pulse-width/pulse-number) or pulse-number modulation to the VTF. Motor distances were adjusted to ensure maximum (\%81.2) localization. Next, the participant matched VTF to the postural sway of the avatar while the computer screen was off. Combined identification accuracy of sway angle and angular velocity was 91\% by only VTF. Finally, the avatar was simulated to cross a road in the VR platform with three conditions (visual on tactile off, visual off tactile on, visual on tactile on) participant. In all conditions, the participant could control the avatar successfully without any falls. Quickest response was obtained when both feedbacks were on (98.7\%), and the worst response was obtained when only VTF was on (92.1\%). VTF seems promising in the proof of concept balance support system presented in this thesis.
Mid-air haptic sensations produced by ultrasound actuators in patients with carpal tunnel syndrome
This thesis utilizes psychophysical experiments with mid-air haptic ultrasound actuators to assess carpal tunnel syndrome (CTS) patients' tactile sensation. 19 female and one male patients (age: 33-61) with unilateral CTS took part in the experiments. We used a two-alternative forced choice task experiments to measure detection thresholds around 250 Hz modulation frequency at the thenar eminence (TE) and at the index finger in affected and healthy hands. In addition, 15 female CTS patients participated in a virtual reality-assisted hand exercise game with haptic feedback. The system usability scale (SUS) and exercise performance scores were evaluated. There was no significant difference in the threshold values from the TE between the CTS hand (M=0.85 au, SD=0.15) and the healthy hand (M=0.87 au, SD=0.16). The thresholds measured from the index fingers of CTS affected hands were all higher than the maximum stimulus level that could be produced by the ultrasonic actuators. For the healthy hands of 17 patients, the detection thresholds were (M=0.90 au, SD=0.09), and the remaining 3 had threshold values above the maximum output of the device. For the exercise game results, there was a significant correlation (ρ = 0.89, p < 0.001) between the SUS (M=80.17%, SD=18.33) and performance scores (M=83.31%, SD=14.80). Since the CTS was at an early stage and there may be a branching (palmar cutaneous branch) of the nerve before entering the carpal tunnel, thresholds were found similar in both hands at the TE. Unfortunately, the limitations of the haptic device did not allow a comparison between index fingers. Moreover, this device and novel technology may be used for the follow-up of the rehabilitation and the treatment of the CTS. As such, the usability of the system is above the criterion value, and in the future it can be improved not to be affected from the performance in the exercise games.
Psychophysical evaluation of a sensory feedback system for prosthetic hands
In this study, a vibrotactile sensory feedback system was designed and tested in accordance with the discrete event-driven sensory feedback control paradigm. Novel approaches were applied in terms of data processing and psychophysical characterization. As the first part, the sensing and signal processing system was designed. Therefore, a robotic hand was equipped with force and bend sensors by mimicking receptors in human hand. The sensor data was recorded during a cylindrical grasping task, and classified for object type and movement phase. Among three machine learning algorithms (k-Nearest Neighbour, Multinomial Logistic Regression and Support Vector Machines), highest classification accuracy was obtained with k-nearest neighbor classifier and the results were promising for the subsequent work. In the second part, the sensory feedback system was designed using two vibrotactile actuators and a user-specific calibration method was presented. The actuators were placed on the upper arms of 10 able-bodied participants. A psychophysical characterization procedure was applied to determine the stimulation amplitudes for each participant specifically. Then, same-different discrimination and pattern recognition experiments were conducted to evaluate the discrimination and closed-set identification of stimuli with varying parameters. Finally, discrete-event driven feedback experiments were run by mapping the parameters of the stimuli to the discrete events related to class labels representing object/movement type. According to the results, the psychophysical characterization procedure was reliable. On the other hand, the performance in the complex tasks was not affected by the psychophysical variations across participants. Experimental results showed that the system can be used to provide object-type and movement-type related information in daily use of prosthetic devices.