Hosseini‐Farid, Mohammad

Abstract Magnetic resonance elastography (MRE) is commonly used as an image-based alternative for palpation of the internal organs of human body. The presence of tumor or other kind of pathologies in biological tissues can increase its stiffness. Therefore, while MRE technique is capable to provide a quantitative measurement, the qualitative description of the tissue stiffness could be potentially informative as well for physicians. MRE can be divided into several steps including the generation of waves in the tissue, measuring the field displacement of the tissue by magnetic resonance imaging devices, and then applying the constitutive based inversion algorithms to measure the material properties of the tissue. The inversion algorithms are dependent to the constitutive model in use, and moreover, it could be computationally expensive. To overcome this hindrance, in this paper, we propose a machine learning framework for categorizing the tissue stiffness based on the magnetic resonance elastography finite element simulation data. In our finite element simulation, the shear waves are generated in an axisymmetrical model by applying harmonic displacement at the center of the model with the known excitation frequency. To obtain the field displacement of the model, in the first step, the natural frequencies of the system will be calculated through numerical Block-Lanczos eigensolver algorithm. Thereafter, a transient dynamic modal analysis is carried out to find the corresponding displacement response of the tissue in different time steps of the simulation. To obtain the training dataset, ten simulations with the pre-assigned linear elastic modulus in the range of 2 to 6 kPa is conducted and the displacement of the tissue in three points at the end of the first and second cycle will be recorded as the features of the dataset. Each instance of the dataset is labelled as “Low“ or “High”, corresponding to its stiffness quantitative value lying in ranges of 2–4 kPa or 4–6 kPa. A machine learning classifying algorithm, a logistic regression hypothesis will be trained on this dataset. The trained hypothesis will be then tested on six new unseen simulation data with known elastic modulus values. The trained logistic regression was able to classify the tissue stiffness with the perfect accuracy score of 1.0. The findings of this study can be used for qualitative description of the tissue stiffness that can be beneficial for pathology diagnosis and moreover, it eliminates the need on the usage of inversion algorithms which leads to reduction in the computational complexity of tissue characterization.

ObjectivesThe mobility (transverse movement) of the median nerve (MN) is decreased in patients with carpal tunnel syndrome and can be measured noninvasively by ultrasound. To date, there are few prognostic features to help predict the outcome of 2 commonly performed treatments: surgical carpal tunnel release and corticosteroid injection. This study aimed to assess the changes in nerve mobility after the intervention and to correlate this with treatment and the disease severity.MethodsA total of 181 patients with carpal tunnel syndrome with different electrophysiologic severities were recruited and assessed by dynamic ultrasound scanning of the MN before and after treatment. The dynamic ultrasound images were collected while the patients performed finger and wrist flexion.ResultsFor both injection and carpal tunnel release, the nerve displacement increased with wrist flexion, from a mean ± SD of 7.0 ± 2.4 to 7.9 ± 2.7 mm (P < .005). Patients who underwent surgery showed greater improvement (P < .005) in nerve mobility compared to those who underwent injection. We also observed that the increase in nerve mobility was predominantly in patients with more nerve damage at baseline.ConclusionsThis study shows that the dynamic behavior of the MN changes in response to treatment and lays a foundation for future studies to assess the prognostic potential of nerve mobility measurement.

Knowing the precise material properties of intracranial head organs is crucial for studying the biomechanics of head injury. It has been shown that these biological tissues are significantly rate-dependent; hence, their material properties should be determined with respect to the range of deformation rate they experience. In this paper, a validated finite element human head model is used to investigate the biomechanics of the head in impact and blast, leading to traumatic brain injuries (TBI). We simulate the head under various directions and velocities of impacts, as well as helmeted and unhelmeted head under blast shock waves. It is demonstrated that the strain rates for the brain are in the range of 36 to 241 s−1, approximately 1.9 and 0.86 times the resulting head acceleration under impacts and blast scenarios, respectively. The skull was found to experience a rate in the range of 14 to 182 s−1, approximately 0.7 and 0.43 times the head acceleration corresponding to impact and blast cases. The results of these incident simulations indicate that the strain rates for brainstem and dura mater are respectively in the range of 15 to 338 and 8 to 149 s−1. These findings provide a good insight into characterizing the brain tissue, cranial bone, brainstem and dura mater, and also selecting material properties in advance for computational dynamical studies of the human head.