Radiological longitudinal follow-up of tumors in CT scans is essential for disease assessment and liver tumor therapy. Currently, most tumor size measurements follow the RECIST guidelines, which can be off by as much as 50%. True volumetric measurements are more accurate but require manual delineation, which is time-consuming and user-dependent. We present a convolutional neural networks (CNN) based method for robust automatic liver tumor delineation in longitudinal CT studies that uses both global and patient specific CNNs trained on a small database of delineated images. The inputs are the baseline scan and the tumor delineation, a follow-up scan, and a liver tumor global CNN voxel classifier built from radiologist-validated liver tumor delineations. The outputs are the tumor delineations in the follow-up CT scan. The baseline scan tumor delineation serves as a high-quality prior for the tumor characterization in the follow-up scans. It is used to evaluate the global CNN performance on the new case and to reliably predict failures of the global CNN on the follow-up scan. High-scoring cases are segmented with a global CNN; low-scoring cases, which are predicted to be failures of the global CNN, are segmented with a patient-specific CNN built from the baseline scan. Our experimental results on 222 tumors from 31 patients yield an average overlap error of 17% (std?=?11.2) and surface distance of 2.1 mm (std?=?1.8), far better than stand-alone segmentation. Importantly, the robustness of our method improved from 67% for stand-alone global CNN segmentation to 100%. Unlike other medical imaging deep learning approaches, which require large annotated training datasets, our method exploits the follow-up framework to yield accurate tumor tracking and failure detection and correction with a small training dataset.
Graphical abstract Flow diagram of the proposed method. In the offline mode (orange), a global CNN is trained as a voxel classifier to segment liver tumor as in [31]. The online mode (blue) is used for each new case. The input is baseline scan with delineation and the follow-up CT scan to be segmented. The main novelty is the ability to predict failures by trying the system on the baseline scan and the ability to correct them using the patient-specific CNN
Atherosclerosis is a type of cardiovascular disease which may cause stroke. It is due to the deposition of fatty plaque in the artery walls resulting in the reduction of elasticity gradually and hence restricting the blood flow to the heart. Hence, an early prediction of carotid plaque deposition is important, as it can save lives. This paper proposes a novel data mining framework for the assessment of atherosclerosis in its early stage using ultrasound images. In this work, we are using 1353 symptomatic and 420 asymptomatic carotid plaque ultrasound images. Our proposed method classifies the symptomatic and asymptomatic carotid plaques using bidimensional empirical mode decomposition (BEMD) and entropy features. The unbalanced data samples are compensated using adaptive synthetic sampling (ADASYN), and the developed method yielded a promising accuracy of 91.43%, sensitivity of 97.26%, and specificity of 83.22% using fourteen features. Hence, the proposed method can be used as an assisting tool during the regular screening of carotid arteries in hospitals.
Prediction of sudden cardiac death continues to gain universal attention as a promising approach to saving millions of lives threatened by sudden cardiac death (SCD). This study attempts to promote the literature from mere feature extraction analysis to developing strategies for manipulating the extracted features to target improvement of classification accuracy. To this end, a novel approach to local feature subset selection is applied using meticulous methodologies developed in previous studies of this team for extracting features from non-linear, time-frequency, and classical processes. We are therefore enabled to select features that differ from one another in each 1-min interval before the incident. Using the proposed algorithm, SCD can be predicted 12 min before the onset; thus, more propitious results are achieved. Additionally, through defining a utility function and employing statistical analysis, the alarm threshold has effectively been determined as 83%. Having selected the best combination of features, the two classes are classified using the multilayer perceptron (MLP) classifier. The most effective features would subsequently be discussed considering their prevalence in the rank-based selection. The results indicate the significant capacity of the proposed method for predicting SCD as well as selecting the appropriate processing method at any time before the incident.
In order to investigate the influence of cusp reduction, cavity isthmus width, and restorative material on stress values in premolar with mesio-occlusal-distal (MOD) cavity, numerical simulations were done on three-dimensional (3D) models of a maxillary second premolar designed using computerized tomography (CT) scan images. The use of four restorative materials (direct resin composite, direct resin composite with resin-modified glass-ionomer cement as the base, indirect resin composite, ceramic), three cavity preparation designs (without cusp coverage, 2-mm palatal cusp coverage, 2-mm palatal and buccal cusp coverage), and two cavity isthmus widths (1/2 and 2/3 intercuspal width) were simulated. After applying a static load of 200 N on the occlusal surface of the tooth, von Mises stresses in the enamel, dentin, and restoration were calculated using finite element analysis (FEA). Stress values in the enamel were primarily influenced by cavity preparation design, while restorative material showed higher contribution in dentin. The lowest stress values were obtained in models with cusp coverage and indirect restorations. Cavity isthmus width had minimal influence on stress values in tooth structures. None of the investigated factors determined stress values in the restoration. In conclusion, the use of ceramic restoration covering both palatal and buccal cusp provided the most favourable stress distribution of premolars with MOD cavity.
Finite element models in conjunction with adequate constitutive relations are pivotal in several physiological and medical applications related to both native and engineered tissues, allowing to predict the tissue response under various loading states. In order to get reliable results, however, the validation of the constitutive models is crucial. Therefore, the main purpose of this work is to provide an experimental-computational approach to the biomechanical investigation of soft tissues such as the dermis. This is accomplished by implementing and validating three widely adopted hyperelastic constitutive models (the Ogden, the Holzapfel, and the Gasser-Ogden-Holzapfel laws) supposed to be adequate to reproduce human reticular dermis mechanical behavior. Biaxial experimental data have represented the basis for the determination of the respective material parameters identified thanks to the definition of a cost function accounting for the discrepancy between experimental and predicted data. Afterwards, the experimental tests have been reproduced through finite element simulations. Hence, the constitutive laws have been validated comparing experimental and numerical outcomes in terms of displacements of four reference points and stress-strain relations. Hence, an experimental-numerical framework is proposed for the investigation of collagenous tissues, which could become more accurate with larger and independent experimental datasets.
Identification of in vivo passive biomechanical properties of healthy human myocardium from regular clinical data is essential for subject-specific modelling of left ventricle (LV). In this work, myocardium was defined by Holzapfel-Ogden constitutive law. Therefore, the objectives of the study were (a) to estimate the ranges of the constitutive parameters for healthy human myocardium using non-invasive routine clinical data, and (b) to investigate the effect of geometry, LV end-diastolic pressure (EDP) and fibre orientations on estimated values. In order to avoid invasive measurements and additional scans, LV cavity volume, measured from routine MRI, and empirical pressure-normalised-volume relation (Klotz-curve) were used as clinical data. Finite element modelling, response surface method and genetic algorithm were used to inversely estimate the constitutive parameters. Due to the ill-posed nature of the inverse optimisation problem, the myocardial properties was extracted by identifying the ranges of the parameters, instead of finding unique values. Additional sensitivity studies were carried out to identify the effect of LV EDP, fibre orientation and geometry on estimated parameters. Although uniqueness of the solution cannot be achieved, the normal ranges of the parameters produced similar mechanical responses within the physiological ranges. These information could be used in future computational studies for designing heart failure treatments.
Electromyography (EMG) in a bio-driven system is used as a control signal, for driving a hand prosthesis or other wearable assistive devices. Processing to get informative drive signals involves three main modules: preprocessing, dimensionality reduction, and classification. This paper proposes a system for classifying a six-channel EMG signal from 14 finger movements. A feature vector of 66 elements was determined from the six-channel EMG signal for each finger movement. Subsequently, various feature extraction techniques and classifiers were tested and evaluated. We compared the performance of six feature extraction techniques, namely principal component analysis (PCA), linear discriminant analysis (LDA), uncorrelated linear discriminant analysis (ULDA), orthogonal fuzzy neighborhood discriminant analysis (OFNDA), spectral regression linear discriminant analysis (SRLDA), and spectral regression extreme learning machine (SRELM). In addition, we also evaluated the performance of seven classifiers consisting of support vector machine (SVM), linear classifier (LC), naive Bayes (NB), k-nearest neighbors (KNN), radial basis function extreme learning machine (RBF-ELM), adaptive wavelet extreme learning machine (AW-ELM), and neural network (NN). The results showed that the combination of SRELM as the feature extraction technique and NN as the classifier yielded the best classification accuracy of 99%, which was significantly higher than those from the other combinations tested.
An understanding of athlete ground reaction forces and moments (GRF/Ms) facilitates the biomechanist’s downstream calculation of net joint forces and moments, and associated injury risk. Historically, force platforms used to collect kinetic data are housed within laboratory settings and are not suitable for field-based installation. Given that Newton’s Second Law clearly describes the relationship between a body’s mass, acceleration, and resultant force, is it possible that marker-based motion capture can represent these parameters sufficiently enough to estimate GRF/Ms, and thereby minimize our reliance on surface embedded force platforms? Specifically, can we successfully use partial least squares (PLS) regression to learn the relationship between motion capture and GRF/Ms data? In total, we analyzed 11 PLS methods and achieved average correlation coefficients of 0.9804 for GRFs and 0.9143 for GRMs. Our results demonstrate the feasibility of predicting accurate GRF/Ms from raw motion capture trajectories in real-time, overcoming what has been a significant barrier to non-invasive collection of such data. In applied biomechanics research, this outcome has the potential to revolutionize athlete performance enhancement and injury prevention.
Finite element (FE) models are increasingly used to validate experimental data in breast cancer. This research constructed a biomechanical FE model for breast shaped phantoms used to develop and validate a mechanical vibration based screening system. Such models do not currently exist but would enhance development of this screening technology. Three phantoms were modelled: healthy, with 10 and 20 mm inclusions. The overall goal was to create models with enough accuracy to replace experimental phantoms in providing data to optimize diagnostic algorithms for digital image-based elasto-tomography (DIET) screening technologies. FE model results were validating against experimental DIET phantom data for over 4000 collected points on each model and phantom using cross-correlation coefficients between experimental simulated data and direct comparison. Results showed good to strong correlation ranging from 0.7 to 1.0 in all cases with over 90% having a value over 0.9. Magnitudes for each frame of the dynamic response also matched well, indicating that the material properties and geometry were accurate enough to provide this level of correlation. These results justify the use of FE model generated data for in silico diagnostic algorithm development testing. The overall modelling and validation approach is not overly complex, and thus generalizable to similar problems using mechanical properties of silicone phantoms, and might be extensible to human cases with further work.
Graphical abstract Validate that dynamic displacements show that the model can be used in place of phantoms for rapid development of diagnostic algorithms that use surface motion to detect underlying mechanical properties.
A 3D convolution neural network (CNN) of deep learning architecture is supplied with essential visual features to accurately classify and segment granulation, necrotic eschar, and slough tissues in pressure ulcer color images. After finding a region of interest (ROI), the features are extracted from both the original and convolved with a pre-selected Gaussian kernel 3D HSI images, combined with first-order models of current and prior visual appearance. The models approximate empirical marginal probability distributions of voxel-wise signals with linear combinations of discrete Gaussians (LCDG). The framework was trained and tested on 193 color pressure ulcer images. The classification accuracy and robustness were evaluated using the Dice similarity coefficient (DSC), the percentage area distance (PAD), and the area under the ROC curve (AUC). The obtained preliminary DSC of 92%, PAD of 13%, and AUC of 95% are promising.
Multidirectional defibrillation protocols have shown better efficiency than monodirectional; still, no testing was performed to assess cell lethality. We investigated lethality of multidirectional defibrillator-like shocks on isolated cardiomyocytes. Cells were isolated from adult male Wistar rats and plated into a perfusion chamber. Electrical field stimulation threshold (ET) was obtained, and cells were paced with suprathreshold bipolar electrical field (E) pulses. Either one monodirectional high-intensity electrical field (HEF) pulse aligned at 0° (group Mono0) or 60° (group Mono60) to cell major axis or a multidirectional sequence of three HEF pulses aligned at 0°, 60°, and 120° each was applied. If cell recovered from shock, pacing was resumed, and a higher amplitude HEF, proportional to ET, was applied. The sequence was repeated until cell death. Lethality curves were built by means of survival analysis from sub-lethal and lethal E. Non-linear fit was performed, and E values corresponding to 50% probability of lethality (E50) were compared. Multidirectional groups presented lethality curves similar to Mono0. Mono60 displayed the highest E50. The novel data endorse the idea of multidirectional stimuli being safer because their effects on lethality of individual cells were equal to a single monodirectional stimulus, while their defibrillatory threshold is lower.
Graphical abstract Monodirectional and multidirectional lethality protocol comparison on isolated rat cardiomyocytes. The heart image is a derivative of “3D Heart in zBrush” (https://vimeo.com/65568770) by Laloxl, used under CC BY 3.0 (https://creativecommons.org/licenses/by/3.0/legalcode)/image extracted from original video.
Vascular morphology characterization is useful for disease diagnosis, risk stratification, treatment planning, and prediction of treatment durability. To quantify the dynamic surface geometry of tubular-shaped anatomic structures, we propose a simple, rigorous Lagrangian cylindrical coordinate system to monitor well-defined surface points. Specifically, the proposed system enables quantification of surface curvature and cross-sectional eccentricity. Using idealized software phantom examples, we validate the method’s ability to accurately quantify longitudinal and circumferential surface curvature, as well as eccentricity and orientation of eccentricity. We then apply the method to several medical imaging data sets of human vascular structures to exemplify the utility of this coordinate system for analyzing morphology and dynamic geometric changes in blood vessels throughout the body.
To improve the compression rates for lossless compression of medical images, an efficient algorithm, based on irregular segmentation and region-based prediction, is proposed in this paper. Considering that the first step of a region-based compression algorithm is segmentation, this paper proposes a hybrid method by combining geometry-adaptive partitioning and quadtree partitioning to achieve adaptive irregular segmentation for medical images. Then, least square (LS)-based predictors are adaptively designed for each region (regular subblock or irregular subregion). The proposed adaptive algorithm not only exploits spatial correlation between pixels but it utilizes local structure similarity, resulting in efficient compression performance. Experimental results show that the average compression performance of the proposed algorithm is 10.48, 4.86, 3.58, and 0.10% better than that of JPEG 2000, CALIC, EDP, and JPEG-LS, respectively.
Bone cells sense mechanical load, which is essential for bone growth and remodeling. In a fracture, this mechanism is compromised. Electromagnetic stimulation has been widely used to assist in bone healing, but the underlying mechanisms are largely unknown. A recent hypothesis suggests that electromagnetic stimulation could influence tissue biomechanics; however, a detailed quantitative understanding of EM-induced biomechanical changes in the bone is unavailable. This paper used a muscle/bone model to study the biomechanics of the bone under EM exposure. Due to the dielectric properties of the muscle/bone interface, a time-varying magnetic field can generate both compressing and shear stresses on the bone surface, where many mechanical sensing cells are available for cellular mechanotransduction. I calculated these stresses and found that the shear stress is significantly greater than the compressing stress. Detailed parametric analysis suggests that both the compressing and shear stresses are dependent on the geometrical and electrical properties of the muscle and the bone. These stresses are also functions of the orientation of the coil and the frequency of the magnetic field. It is speculated that the EM field could apply biomechanical influence to fractured bone, through the fine-tuning of the controllable field parameters.
Acute limp is a common presenting condition in the paediatric emergency department. There are a number of causes of acute limp that include traumatic injury, infection and malignancy. These causes in young children are not easily distinguished. In this pilot study, an infrared thermographic imaging technique to diagnose acute undifferentiated limp in young children was developed. Following required ethics approval, 30 children (mean age =?5.2 years, standard deviation =?3.3 years) were recruited. The exposed lower limbs of participants were imaged using a high-resolution thermal camera. Using predefined regions of interest (ROI), any skin surface temperature difference between the healthy and affected legs was statistically analysed, with the aim of identifying limp. In all examined ROIs, the median skin surface temperature for the affected limb was higher than that of the healthy limb. The small sample size recruited for each group, however, meant that the statistical tests of significant difference need to be interpreted in this context. Thermal imaging showed potential in helping with the diagnosis of acute limp in children. Repeating a similar study with a larger sample size will be beneficial to establish reproducibility of the results.
Graphical abstract A young child with an acute undifferentiated limp undergoes thermal imaging and the follow on image analysis assists the limp diagnosis.
An atlas-based multimodal registration method for 2-dimension images with discrepancy structures was proposed in this paper. Atlas was utilized for complementing the discrepancy structure information in multimodal medical images. The scheme includes three steps: floating image to atlas registration, atlas to reference image registration, and field-based deformation. To evaluate the performance, a frame model, a brain model, and clinical images were employed in registration experiments. We measured the registration performance by the squared sum of intensity differences. Results indicate that this method is robust and performs better than the direct registration for multimodal images with discrepancy structures. We conclude that the proposed method is suitable for multimodal images with discrepancy structures.
The conductivity of brain tissues is not only essential for electromagnetic source estimation (ESI), but also a key reflector of the brain functional changes. Different from the other brain tissues, the conductivity of whiter matter (WM) is highly anisotropic and a tensor is needed to describe it. The traditional electrical property imaging methods, such as electrical impedance tomography (EIT) and magnetic resonance electrical impedance tomography (MREIT), usually fail to image the anisotropic conductivity tensor of WM with high spatial resolution. The diffusion tensor imaging (DTI) is a newly developed technique that can fulfill this purpose. This paper reviews the existing anisotropic conductivity models of WM based on the DTI and discusses their advantages and disadvantages, as well as identifies opportunities for future research on this subject. It is crucial to obtain the linear conversion coefficient between the eigenvalues of anisotropic conductivity tensor and diffusion tensor, since they share the same eigenvectors. We conclude that the electrochemical model is suitable for ESI analysis because the conversion coefficient can be directly obtained from the concentration of ions in extracellular liquid and that the volume fraction model is appropriate to study the influence of WM structural changes on electrical conductivity.
Respiratory inductance plethysmography (RIP) sensor is an inexpensive, non-invasive, easy-to-use transducer for collecting respiratory movement data. Studies have reported that the RIP signal’s amplitude and frequency can be used to discriminate respiratory diseases. However, with the conventional approach of RIP data analysis, respiratory muscle effort cannot be estimated. In this paper, the estimation of the respiratory muscle effort through RIP signal was proposed. A complementary ensemble empirical mode decomposition method was used, to extract hidden signals from the RIP signals based on the frequency bands of the activities of different respiratory muscles. To validate the proposed method, an experiment to collect subjects’ RIP signal under thoracic breathing (TB) and abdominal breathing (AB) was conducted. The experimental results for both the TB and AB indicate that the proposed method can be used to loosely estimate the activities of thoracic muscles, abdominal muscles, and diaphragm.
Vertebral fractures are common in children with osteogenesis imperfecta (OI). Current imaging methods for fracture detection (X-ray and DXA) use ionising radiation. This pilot study explored whether the alteration in blood flow in vertebral fractures results in skin temperature changes that may be detected using high resolution thermal imaging (HRTI) and thus assist diagnosis and monitoring of fractures in OI patients. Eleven participants aged 5–18 years with OI and known vertebral fractures were enrolled. Small metal discs were placed on the skin surface alongside the vertebrae before participants had DXA and X-ray scans and thermal imaging of their backs. Visibility of the discs on the DXA and X-ray scans and thermal images allowed the temperatures of the skin surface above vertebrae without (healthy) and with fractures to be compared to their respective adjacent skin surface regions (region of reference, ROR) by calculating the temperature percentage change (TPC). The TPC between the skin temperature over the fractured thoracic vertebrae (n?=?11) and the ROR was significant (1.44%, p?=?0.002, 95% confidence). TPC between the skin temperature over healthy thoracic vertebrae and ROR was not significant (0.97%, p?=?0.15, 95% confidence). HRTI may provide a novel tool for assisting in detection of vertebral fractures in OI.
Graphical abstract ? Patients (aged 5–18) with osteogenesis imperfecta and known vertebral fractures. ? Thermal imaging was performed alongside routine imaging (DXA scan and spinal X-ray). ? The temperature above each vertebra was compared with its adjacent skin region to assist with diagnosis of the fracture.
