A probabilistic approach to tomography and adjoint state methods, with an application to full waveform inversion in medical ultrasound

Healthy human skin Kelvin-Voigt fractional and spring-pot biomarkers reconstruction using torsional wave elastography

This paper presents a novel method for reconstructing skin parameters using Probabilistic Inverse Problem (PIP) techniques and Torsional Wave Elastography (TWE) rheological modeling. A comprehensive examination was conducted to compare and analyze the theoretical, time-of-flight (TOF), and full-signal waveform (FSW) approaches. The objective was the identification of the most effective method for the estimation of mechanical parameters. Initially, the most appropriate rheological model for the simulation of skin tissue behavior was determined through the application and comparison of two models, spring pot (SP) and Kevin Voigt fractional derivative (KVFD). A numerical model was developed using the chosen rheological models. The collection of experimental data from 15 volunteers utilizing a TWE sensor was crucial for obtaining significant information for the reconstruction process. The study sample consisted of five male and ten female subjects ranging in age from 25 to 60 years. The procedure was performed on the ventral forearm region of the participants. The process of reconstructing skin tissue parameters was carried out using PIP techniques. The experimental findings were compared with the numerical results. The three methods considered (theoretical, TOF, FSW) have been used. The efficacy of TOF and FSW was then compared with theoretical method. The findings of the study demonstrate that the FSW and TOF techniques successfully reconstructed the parameters of the skin tissue in all of the models. The SP model's the skin tissue η values ranged from 8 to 12 P a · s , as indicated by the TOF reconstruction parameters. η values found by the KVFD model ranged from 4.1 to 9.3 P a · s . The μ values generated by the KVFD model range between 0.61 and 96.86 kPa. However, FSW parameters reveal that skin tissue η values for the SP model ranged from 7.8 to 12 P a · s . The KVFD model determined η values between 6.3 and 9.5 P a · s . The KVFD model presents μ values ranging between 26.02 and 122.19 kPa. It is shown that the rheological model that best describes the nature of the skin is the SP model and its simplicity as it requires only two parameters, in contrast to the three parameters required by the KVFD model. Therefore, this work provides a valuable addition to the area of dermatology, with possible implications for clinical practice.

Comparing the performance of machine learning methods in estimating the shear wave transit time in one of the reservoirs in southwest of Iran

Shear wave transit time is a crucial parameter in petroleum engineering and geomechanical modeling with significant implications for reservoir performance and rock behavior prediction. Without accurate shear wave velocity information, geomechanical models are unable to fully characterize reservoir rock behavior, impacting operations such as hydraulic fracturing, production planning, and well stimulation. While traditional direct measurement methods are accurate but resource-intensive, indirect methods utilizing seismic and petrophysical data, as well as artificial intelligence algorithms, offer viable alternatives for shear wave velocity estimation. Machine learning algorithms have been proposed to predict shear wave velocity. However, until now, a comprehensive comparison has not been made on the common methods of machine learning that had an acceptable performance in previous researches. This research focuses on the prediction of shear wave transit time using prevalent machine learning techniques, along with a comparative analysis of these methods. To predict this parameter, various input features have been employed: compressional wave transit time, density, porosity, depth, Caliper log, and Gamma-ray log. Among the employed methods, the random forest approach demonstrated the most favorable performance, yielding R-squared and RMSE values of 0.9495 and 9.4567, respectively. Furthermore, the artificial neural network, LSBoost, Bayesian, multivariate regression, and support vector machine techniques achieved R-squared values of 0.878, 0.8583, 0.8471, 0.847 and 0.7975, RMSE values of 22.4068, 27.8158, 28.0138, 28.0240 and 37.5822, respectively. Estimation analysis confirmed the statistical reliability of the Random Forest model. The formulated strategies offer a promising framework applicable to shear wave velocity estimation in carbonate reservoirs.

Brain tumour genetic network signatures of survival

Tumour heterogeneity is increasingly recognized as a major obstacle to therapeutic success across neuro-oncology. Gliomas are characterized by distinct combinations of genetic and epigenetic alterations, resulting in complex interactions across multiple molecular pathways. Predicting disease evolution and prescribing individually optimal treatment requires statistical models complex enough to capture the intricate (epi)genetic structure underpinning oncogenesis. Here, we formalize this task as the inference of distinct patterns of connectivity within hierarchical latent representations of genetic networks. Evaluating multi-institutional clinical, genetic and outcome data from 4023 glioma patients over 14 years, across 12 countries, we employ Bayesian generative stochastic block modelling to reveal a hierarchical network structure of tumour genetics spanning molecularly confirmed glioblastoma, IDH-wildtype; oligodendroglioma, IDH-mutant and 1p/19q codeleted; and astrocytoma, IDH-mutant. Our findings illuminate the complex dependence between features across the genetic landscape of brain tumours and show that generative network models reveal distinct signatures of survival with better prognostic fidelity than current gold standard diagnostic categories.

3D Ultrasonic Brain Imaging with Deep Learning Based on Fully Convolutional Networks

Compared to magnetic resonance imaging (MRI) and X-ray computed tomography (CT), ultrasound imaging is safer, faster, and more widely applicable. However, the use of conventional ultrasound in transcranial brain imaging for adults is predominantly hindered by the high acoustic impedance contrast between the skull and soft tissue. This study introduces a 3D AI algorithm, Brain Imaging Full Convolution Network (BIFCN), combining waveform modeling and deep learning for precise brain ultrasound reconstruction. We constructed a network comprising one input layer, four convolution layers, and one pooling layer to train our algorithm. In the simulation experiment, the Pearson correlation coefficient between the reconstructed and true images was exceptionally high. In the laboratory, the results showed a slightly lower but still impressive coincidence degree for 3D reconstruction, with pure water serving as the initial model and no prior information required. The 3D network can be trained in 8 h, and 10 samples can be reconstructed in just 12.67 s. The proposed 3D BIFCN algorithm provides a highly accurate and efficient solution for mapping wavefield frequency domain data to 3D brain models, enabling fast and precise brain tissue imaging. Moreover, the frequency shift phenomenon of blood may become a hallmark of BIFCN learning, offering valuable quantitative information for whole-brain blood imaging.

A Forward Model Incorporating Elevation-Focused Transducer Properties for 3-D Full-Waveform Inversion in Ultrasound Computed Tomography

Ultrasound computed tomography (USCT) is an emerging medical imaging modality that holds great promise for improving human health. Full-waveform inversion (FWI)-based image reconstruction methods account for the relevant wave physics to produce high spatial resolution images of the acoustic properties of the breast tissues. A practical USCT design employs a circular ring-array comprised of elevation-focused ultrasonic transducers, and volumetric imaging is achieved by translating the ring-array orthogonally to the imaging plane. In commonly deployed slice-by-slice (SBS) reconstruction approaches, the 3-D volume is reconstructed by stacking together 2-D images reconstructed for each position of the ring-array. A limitation of the SBS reconstruction approach is that it does not account for 3-D wave propagation physics and the focusing properties of the transducers, which can result in significant image artifacts and inaccuracies. To perform 3-D image reconstruction when elevation-focused transducers are employed, a numerical description of the focusing properties of the transducers should be included in the forward model. To address this, a 3-D computational model of an elevation-focused transducer is developed to enable 3-D FWI-based reconstruction methods to be deployed in ring-array-based USCT. The focusing is achieved by applying a spatially varying temporal delay to the ultrasound pulse (emitter mode) and recorded signal (receiver mode). The proposed numerical transducer model is quantitatively validated and employed in computer simulation studies that demonstrate its use in image reconstruction for ring-array USCT.

Transcranial Ultrasound Imaging With Decomposition Descent Learning-Based Full Waveform Inversion

Noninvasive brain diagnosis is extremely important because of its efficiency, low cost, and painless nature in the prediction of stroke, cerebral hemorrhage, and other brain research. At present, achieving full 3-D quantitative ultrasonic imaging of the human brain is a cutting-edge challenge due to the complex structures of the human brain and the strong scattering caused by the skulls. In this article, we achieved quantitative ultrasonic imaging of inside-brain anomalies with our proposed method, the decomposition descent learning-based full waveform inversion (DDL-FWI). The proposed method adopts a linear residual decomposing technique to greatly alleviate the computation burden in fast inversion tomography (FIT) with enhanced convergence guaranteed by residual functions. Testing results in both simulation and laboratory experiments demonstrated that our method can achieve high-quality quantitative imaging of brain soft tissues and skulls even starting from homogeneous water background in 2-D, and this method is capable of reconstructing both complex brain tissues and clots in 2-D and 3-D cases using either clean or noisy signals, with a robust 3-D clot resolution as small as 18 mm and 2-D reconstruction speed in 11.20 s. Combined with advanced ultrasonic hardware, DDL-FWI can be easily trained and used for brain imaging efficiently that frees patients from harmful influences from traditional imaging techniques, e.g., ionization radiations from X-ray computed tomography (CT).