The aim of this work was to test microwave brain stroke detection and classification using support vector machines (SVMs). We tested how the nature and variability of training data and system parameters impact the achieved classification accuracy. Using experimentally verified numerical models, a large database of synthetic training and test data was created. The models consist of an antenna array surrounding reconfigurable geometrically and dielectrically realistic human head phantoms with virtually inserted strokes of arbitrary size, and different dielectric parameters in different positions. The generated synthetic data sets were used to test four different hypotheses, regarding the appropriate parameters of the training dataset, the appropriate frequency range and the number of frequency points, as well as the level of subject variability to reach the highest SVM classification accuracy. The results indicate that the SVM algorithm is able to detect the presence of the stroke and classify it (i.e., ischemic or hemorrhagic) even when trained with single-frequency data. Moreover, it is shown that data of subjects with smaller strokes appear to be the most suitable for training accurate SVM predictors with high generalization capabilities. Finally, the datasets created for this study are made available to the community for testing and developing their own algorithms.
25 August 2019 in International Journal of Antennas and Propagation
This work presents the design and manufacturing of an anatomically and dielectrically realistic layered phantom of the human head that allows the insertion of ischemic and hemorrhagic stroke phantom models. A 2.5D physical phantom was designed using a representative anatomical image of the human head, which was simplified into 5 different layers that mimic the scalp, skull, cerebrospinal fluid, brain, and stroke regions in terms of anatomy and dielectric properties. Apart from the brain phantom, all other layers consist of a mixture of polyurethane rubber, graphite powder, and carbon black powder. The brain phantom is in the liquid form to facilitate the insertion of different stroke models (ischemic or hemorrhagic) with different positions and shapes. Phantoms were designed with dielectric properties valid within the frequency range 0.5–3.0 GHz, which is relevant for microwave stroke detection and classification. Molds for casting individual parts of the phantom were printed in 3D. The presented phantom is suitable for the development and testing of microwave systems and algorithms used in the detection and classification of vascular events relevant to stroke diagnosis.
Due to the clinically proven benefit of hyperthermia treatments if added to standard cancer therapies for various tumor sites and the recent development of non-invasive temperature measurements using magnetic resonance systems, the hyperthermia community is convinced that it is a time when even patients with brain tumors could benefit from regional microwave hyperthermia, even if they are the subject of a treatment to a vital organ. The purpose of this study was to numerically analyze the ability to achieve a therapeutically relevant constructive superposition of electromagnetic (EM) waves in the treatment of hyperthermia targets within the brain. We evaluated the effect of the target size and position, operating frequency, and the number of antenna elements forming the phased array applicator on the treatment quality. In total, 10 anatomically realistic 2D human head models were considered, in which 10 circular hyperthermia targets with diameters of 20, 25, and 30 mm were examined. Additionally, applicators with 8, 12, 16, and 24 antenna elements and operating frequencies of 434, 650, 915, and 1150 MHz, respectively, were analyzed. For all scenarios considered (4800 combinations), the EM field distributions of individual antenna elements were calculated and treatment planning was performed. Their quality was evaluated using parameters applied in clinical practice, i.e., target coverage (TC) and the target to hot-spot quotient (THQ). The 12-antenna phased array system operating at 434 MHz was the best candidate among all tested systems for HT treatments of glioblastoma tumors. The 12 antenna elements met all the requirements to cover the entire target area; an additional increase in the number of antenna elements did not have a significant effect on the treatment quality.
For a first time 20 human blood samples with varying glucose concentrations using previously introduced metamaterial microwave (MTM) sensors, which had been already tested with phantoms. The MTM sensors based on composite right/left-handed metamaterial transmission line showed a detectable phase shift in the forward transmission parameter (∠S21) which is related to the blood glucose concentration. Application of a novel biochemical approach introduces separate mathematical models based on gender, reflecting physiological difference in blood, which were further evidenced by measured variations in blood dielectric properties. This contributed to enhanced sensor performance as well as the professional sample preservation system. The accuracy of glucose concentration estimation was analyzed using mean absolute relative difference (MARD) and Parkes error grids. The MARD values for the Female and Male group reached 23 % and 29 % respectively. The findings demonstrate the potential of the MTM sensor for non-invasive glycemia monitoring.
3 June 2024 in IEEE Journal of Electromagnetics, RF and Microwaves in Medicine and Biology
The primary objective of this study is to systematically evaluate the performance of the Support Vector Machine (SVM) algorithm, identifying optimal configurations and appropriate parameters for training and testing data, for microwave brain stroke classification. Using experimentally verified 3D numerical models, a large database of synthetic training and test data has been created with different levels of data variability. These models consist of an antenna array surrounding reconfigurable geometrically and dielectrically realistic human head models Within these models, strokes of varying sizes, types, and dielectric parameters are virtually inserted at different positions in brain within the plane of the antennas. Synthetic data sets have been generated to study the impact of reducing training data, data dimensionality, data format, and algorithm settings. The results of this study confirm that Principal Component Analysis (PCA) dimensionality reduction significantly improved the classification accuracy of the SVM algorithm, and datasets of subjects with smaller strokes appeared to be the most suitable for training. Furthermore, datasets that contain the real and imaginary parts of transmission and reflection coefficients result in the highest classification accuracy. For the current antenna array, the best observed setting and scenarios with high variability in training and test data, close to real clinical scenarios, the ability to accurately classify ischemic strokes and suggest safe initiation of thrombotic therapy is approximately 70%.
16 April 2025 in IEEE Transactions on Instrumentation and Measurement
Knowledge of the dielectric parameters of biological tissues is important in a wide range of medical applications of electromagnetic fields. Commercial systems using high-quality vector network analyzers (VNA) are expensive and can be a barrier to the development of existing and emerging applications. Therefore, the main objective of this work was to investigate the potential of using inexpensive VNA, which is available on the market, to measure the dielectric parameters of biological tissues. To accomplish this task, a system for the measurement of dielectric properties of biological tissues in the frequency range from 70 MHz to 3 GHz using inexpensive VNA was developed and presented. The system consists of pocketVNA 2.0 and a custom-made coaxial reflection probe with a diameter of 2.2 mm. The virtual line model of the coaxial probe was used to convert the measured reflection coefficients to the complex permittivity. A total of five ex-vivo tissues (heart, muscle, liver, kidney cortex, and fat) and two standard liquids (isopropanol and ethanol) were measured. The lower accuracy and lower signal-to-noise ratio of the inexpensive VNA were compensated for by post-processing of the measured data, including a time-domain gating and fitting of the evaluated dielectric properties to the Cole-Cole and Debye models for tissues and standard liquids, respectively. The measurement results were compared to the results obtained with a commercial measurement system, and both the standard deviation and mean absolute percentage error (MAPE) are evaluated and presented. The lowest achieved MAPE of biological tissues was for fitted relative permittivity 4.35% (fat, f < 200 MHz) and 0.66% (kidney cortex, f > 200 MHz) and for fitted equivalent electrical conductivity 2.58% (heart, f < 200 MHz) and 3.83% (heart, f > 200 MHz).