Machine Learning for Medical Imaging
At Kurtlab, we develop novel computer vision techniques and deep learning models to address important clinical challenges requiring multimodal medical image registration, instance segmentation, and image synthesis. We recently developed a novel architecture for the challenging task of deformable image registration to establish correspondences between pre-operative and follow-up scans of the same patient diagnosed with an adult brain diffuse high-grade glioma. For this task, we developed a cascaded architecture composed of several Inception modules and a variant of TransMorph. The dataset for each patient was comprised of a native pre-contrast (T1), a contrast-enhanced T1-weighted (T1-CE), a T2-weighted (T2), and a Fluid Attenuated Inversion Recovery (FLAIR). The Inception model was used to fuse the 4 image modalities together and extract the most relevant information. Then, a variant of the TransMorph architecture was adapted to generate the displacement fields.
Figure. Example of warped image with output displacement field of the network: a) Moving image, b) target image, c) warped image
Conferences and Publications:
 J. Abderezaei*, A. Pionteck*, A. Chopra, M. Kurt, “3D Inception-Based TransMorph: Pre- and Post-operative Multi-contrast MRI Registration in Brain Tumors”, International MICCAI Brain Tumor Sequence Registration (BraTS-Reg) Challenge, Sept. 18-22, Singapore.
 VIT_FNO: A. Chopra, A. Pionteck, J. Abderezaei, and M. Kurt, “VIT-FNO; A Robust Model For Tracking Motion In 4d-mri,” Summer Biomechanics, Bioengineering, and Biotransport Conference (SB3C), June 20-23 2022, Eastern Shore, MD, USA.
Biomechanics of the Human Brain During Head Impacts
Traumatic brain injury (TBI) is a major health issue affecting millions of people across the world. Considering such a high number of incidents, however, the physical cause of TBI is still unknown. Investigating the patterns of brain motion and deformation after a head impact is one of the novel methods that have recently shed light on this issue. Here at Kurtlab, we use advanced finite element simulations and modal analysis to characterize the nonlinear dynamical behavior of the brain-skull system as a result of concussive head impacts. We are specifically interested in understanding the dynamics of the deep white matter and dural folds which are shown to be important in concussion pathology.
Funded by: National Science Foundation (NSF)
 Abderezaei, J., Zhao, W., Grijalva, C. L., Fabris, G., Ji, S., Laksari, K., & Kurt, M. (2019). Nonlinear dynamical behavior of the deep white matter during head impact. Physical Review Applied, 12(1), 014058.
 Laksari, K., Kurt, M., Babaee, H., Kleiven, S., & Camarillo, D. (2018). Mechanistic insights into human brain impact dynamics through modal analysis. Physical review letters, 120(13), 138101.
Mechanical Characterization of Brain-Skull Interface
The mammalian meninges are a complex and anatomically heterogeneous tissue that surrounds the brain and tethers it to the skull. Characterizing the mechanical properties of the different meningeal layers is crucial in order to be able to predict load propagation to the brain during injury events, for instance via finite element or other types of computer simulations.
At Kurtlab, we aim at experimentally characterizing the coupling between anatomical structure and local mechanical properties of rat and mouse meninges via a combination of techniques that includes atomic force microscopy, fluorescent microscopy, optical coherence tomography, and microindentation.
You can read about some of our initial results here!
 G. Fabris*, Z. Suar*, and M. Kurt. Micromechanical heterogeneity of the rat pia-arachnoid complex, Acta Biomaterialia, doi: 10.1016/j.actbio.2019.09.044, 2019.
 Z. Suar*, G. Fabris*, and M. Kurt. Isolation and immunofluorescent staining of fresh rat pia-arachnoid complex tissue for micromechanical characterization, Current Protocols in Neuroscience, doi: https://doi.org/10.1002/cpns.83, 2019, Featured Cover Article
Effects of Microstructural Variation in TBI Vulnerability
The corpus callosum is one of the regions of the brain that has been often implicated in TBI studies, and it experiences a large strain in mTBI. The brain stiffness, which is correlated with myelin content, has an impact on the injury process. Our goal is to study the effect of corpus callosum stiffness on TBI mechanisms by combining in vivo experiments with computational simulations. We have developed an experimental animal study in which a surgical procedure is used to induce local demyelination in the corpus callosum, followed by a weight drop protocol to cause a mild diffuse injury in the murine brain. We then analyze the patterns of injury and compare the results with impact simulations of the murine brain.
Pulsatile Motion of the Human Brain Through 3D Amplified MRI
Intrinsic brain deformities could provide important clinical insights for several neuropathologies, for example as a new diagnostic marker for Chiari I (CM-I) malformation. However, the main practical clinical difficulty in accessing this information is that brain movement is extremely subtle and therefore difficult to detect. Recently, a new approach has been introduced: amplified magnetic resonance imaging (aMRI) is an MR post-processing technique that retrospectively amplifies cardiac CINE MRI data sets; essentially, it amplifies movement in specific frequency bands, making subtle brain deformities immediately visible. The original MRI approach has so far only been applied to two-dimensional (2D) data sets. This represents a major limitation in the attempt to better understand a range of pathophysiological mechanisms in the brain. Our current work is therefore focused on the development of a 3D aMRI method. The developed 3D aMRI method is based on an improvement of the 2D aMRI algorithm, allowing the visualization and quantification of the volumetric displacements of 3D pixels in 3 directions.
Funded by: NIH NINDS 1R21NS111415 award
Studying Aneurysm Instability through Amplified Flow Imaging (aFlow)
Intracranial aneurysms are localized arterial enlargements that have a high mortality rate upon rupture. Therefore, reliable and quantifiable criteria enabling to monitor aneurysm progression and to predict rupture risk are of high interest in the clinical setting. The current methods to assess its rupture risk, however, are unreliable, since aneurysms that are deemed stable, might rupture unpredictably. Recent studies have suggested a correlation between the patterns of motion in the wall of the intracranial aneurysm with their stability, where aneurysms with abnormal wall motion are more prone to rupture. Here, we have developed a new image processing technique called aFlow, which is an altered version of aMRI, that allows visualizing the fast-changing deformations of the aneurysm wall (higher harmonics). With aFlow, we are analyzing whether these higher harmonics motions could have diagnostic implications.
 Terem, I., Ni, W.W., Goubran, M., Rahimi, M.S., Zaharchuk, G., Yeom, K.W., Moseley, M.E., Kurt, M. and Holdsworth, S.J., 2018. Revealing sub‐voxel motions of brain tissue using phase‐based amplified MRI (aMRI). Magnetic resonance in medicine, 80(6), pp.2549-2559.
 J. Abderezaei, J. Martinez, I. Terem, G. Fabris, S. Holdsworth, K. Nael and M. Kurt, “Amplified Flow Imaging (aFlow): A Novel Technique to Visualize Intracranial Aneurysms Dynamics”, SB3C, July 17-20th 2020, Vail, CO, USA
 J. Martinez, J. Abderezaei, S. Holdsworth, K. Nael, M. Kurt, “aFlow: A Novel Technique for Imaging Cerebrovascular Motions”, BMES, Oct. 17-20th 2018, Atlanta, GA, USA)
In vivo Mechanical Characterization of the Human Brain Through MR Elastography
Magnetic resonance elastography (MRE) has the promise of mapping the viscoelastic properties of the human brain. Viscoelasticity is a measure of the microstructural composition of the soft biological tissue and is increasingly used as a diagnostic marker. MRE has been gaining attention as a non-invasive means to measure the viscoelasticity of tissues in vivo, most prominently brain tissue. MRE obtains information about the stiffness of the tissue by assessing the propagation of mechanical waves through the tissue with a magnetic resonance imaging (MRI) technique. Several studies have shown that it can potentially be used as a diagnostic tool for detecting the onset of neurological diseases including Alzheimer’s and traumatic brain injury.
The characterization of brain tissue mechanical properties is also of crucial importance in the development of realistic numerical models of the human head. Our research aims to incorporate in vivo human brain measurements into brain finite element models.
BioInterface Characterization through MRE
Most tissues consist of multiple layers and membranes, and abnormalities such as tumors, all of which interact through bio-interfaces. Detailed mechanical characterization of bio-interfaces under physiological conditions (ideally in vivo) is essential for accurate diagnosis, prognosis, and effective treatment on biomedical applications. With the help of new MR imaging modalities such as MR elastography (MRE) and amplified MRI (aMRI), now we are capable of investigating the short term temporal variation of strain fields in vivo in the vicinity of bio-interfaces.
Funded By: National Science Foundation CMMI Dynamics, Control and System Diagnostics Program
Biomechanics of Brain Development Through MRE
At KurtLab, we aim to understand the unique mechanical properties of the developing pediatric brain and aging adult brain through MRE. This will enable us to predict the functional and structural tolerance of the human brain to traumatic brain injury, and develop protective equipment for the pediatric population. Specifically, if the susceptible regions of the pediatric brain to trauma over different ages can be determined, this would have a major interventional effect in the form of novel diagnostic/therapeutic/preventive tools. This will also allow us to determine how different regions of the brain change mechanically and therefore microstructurally as children age into adults.
Nonlinear System Identification in Mechanical and Biological Systems
Nonlinear Modal Analysis in the Human Brain
Understanding human brain mechanics associated with TBI represents a crucial step towards improving protective devices and patient outcomes. The brain is a complex biomechanical system with intricate geometry, nonuniform interfacial boundary conditions, and significantly inhomogeneous and nonlinear material properties. Peculiar nonlinear dynamical behaviors have been evidenced in several studies of head impacts using brain models.
We aim to investigate the fundamental nonlinear dynamics of a simplified finite element (FE) brain model. We modeled the brain as a long cylinder with dimensions similar to that of the human brain with a thin membrane embedded in the center. The cylinder is subjected to sudden harmonic rotations applied at its external boundary around the main vertical axis. We will also test this in a similar MRI phantom setup. This 3D printed setup will provide a high deceleration to a brain-mimicking silicone phantom with a thin plastic membrane to represent the falx cerebri.
Empirical System Identification in Nonlinear Systems
Practical engineering applications necessitate system identification techniques for analyses, design, and maintenance. System identification techniques existing in the literature work quite well for linear dynamical systems, e.g. linear modal analyses. However, these techniques fail to produce meaningful results when structures with nonlinear couplings are analyzed. The same holds for biological systems as well: As biological systems become more complex and assumptions of linearity no longer hold, a need for more advanced system identification and reduced-order modeling techniques arises.
Smart Biomedical Devices
Leveraging Smart Mouthguards to Study Soccer Heading Scenarios
Soccer, one of the most popular sports in the world, is unique in that players can field the ball with their head without requiring any head protection. The effect of these repetitive subconcussive impacts, or impacts that result in head accelerations below the threshold to cause mild TBI, is relatively unknown. We hypothesize that the accumulation of these subconcussive impacts will induce measurable neurocognitive and functional changes on the player. To test this, we will ask players to perform a series of headers wearing instrumented mouthguards that measures 6 degrees-of-freedom (DOF) translational and rotational head kinematics. Using this data, we will correlate head kinematics to neurocognitive metrics, and determine if different types of heading (i.e. different impact locations) will result in larger stress and strain on the brain.
Funded by: CHI Healthcare Scholars Program
MRE Actuators for Better Diagnostics and Prognosis of IVD Degeneration
The intervertebral disc (IVD) is a flexible pad located in between each vertebra and acts as a shock absorber for the spine. When the disc experiences high pressure, small tears may occur as well as thinning along the annulus fibrosus. These structural changes lead to the nucleus pulposus drying out, causing a loss of flexibility and thus degeneration. In order to better diagnose these pathological changes in the IVD, an MR compatible device for automated actuation and positioning for magnetic resonance elastography (MRE) is developed.
Collaborator: Jun Ueda
Funded by: Department of Defense Peer-Reviewed Medical Research Program for the Office of the Congressionally Directed Medical Research Programs
Testing and Design of Smart Helmets in Contact Sports
A current particular focus at KurtLab is smart preventive equipment design in sports.
Helmets have proven effective in mitigating moderate to severe head injuries by attenuating translational head accelerations. However, reduced acceleration levels do not seem to prevent mild traumatic brain injury (mTBI). Despite the ubiquity of helmets in contact sports, mTBI is highly prevalent.
With the availability of high-rate MEMS sensors and high performance batteries, a new class of helmets, i.e. smart, expandable helmets, can be designed that can sense an impending collision and expand to protect the brain in an optimized manner. We currently work on developing smart and expandable helmets and neck protection systems for contact sports. In previous work, (Kurt et al, 2016, ABME) we have shown that such helmets can reduce the risk of head injuries by 8-fold compared to conventional sports helmets in football and bicycling.