Geodesic Distances in Probabilistic Spaces for Patch-Based Ultrasound Image Processing.

We propose a novel method to improve the performance of existing three dimensional (3D) human face recognition algorithms that employ Euclidean distances between facial fiducial points as features. We further investigate a novel 3D face recognition algorithm that employs geodesic and Euclidean distances between facial fiducial points. We demonstrate that this algorithm is robust to changes in facial expression. Geodesic and Euclidean distances were calculated between pairs of 25 facial fiducial points. For the proposed algorithm, geodesic distances and 'global curvature' characteristics, defined as the ratio of geodesic to Euclidean distance between a pairs of points, were employed as features. The most discriminatory features were selected using stepwise linear discriminant analysis (LDA). These were projected onto 11 LDA directions, and face models were matched using the Euclidean distance metric. With a gallery set containing one image each of 105 subjects and a probe set containing 663 images of the same subjects, the algorithm produced EER=1.4% and a rank 1 RR=98.64%. It performed significantly better than existing algorithms based on principal component analysis and LDA applied to face range images. Its verification performance for expressive faces was also significantly better than an algorithm that employed Euclidean distances between facial fiducial points as features.

In spite of the advantages of ease of imaging and low acquisition cost, ultrasound images are noisier and have poorer image quality than other imaging modalities like CT or MRI and hence require an experienced clinician for interpretation. Processing by automated segmentation assists clinicians and improves the accuracy of assessment by minimizing its subjective nature. Various methods for the segmentation of ultrasound images exist, but there is not much literature on the processing of 2D ultrasound images of the foot. This work aimed at developing a novel method for image processing in order to achieve automated segmentation of ultrasound images of the plantar soft tissue. Preprocessing of the ultrasound images was performed using the anisotropic diffusion filter followed by contrast enhancement. The Chan-Vese active contour method was used for segmentation. Our method took into account the difficulty of visualization of the tissues and bony structures in the foot and used an additional curvature parameter for segmentation. Assessing the changes in the biomechanical properties of the plantar soft tissue can be a potential application of this method especially in case of the diabetic foot.

Cross-modal Transfer Learning Based on an Improved CycleGAN Model for Accurate Kidney Segmentation in Ultrasound Images

External ventricular drains (EVD) are widely used to measure and manage intracranial pressure (ICP) for aneurysmal subarachnoid hemorrhage (aSAH) patients. After several days of use a decision is made to remove the drain or replace it with an indwelling shunt. This involves a “clamping trial” whereby the EVD is clamped and transduced and pre and post clamping CT imaging is performed to observe any change in ventricular size and clinical status. This practice may lead to prolonged hospital stay, extra radiation exposure, and neurological insult. The present study aims to apply a widely validated morphological clustering analysis of ICP pulse (MOCAIP) algorithm to detect signatures from the pulse waveform to differentiate an intact CSF circulatory system from an abnormal one during EVD clamping. We hypothesize that an intact CSF system should be stable and pulses with a similar mean ICP level should have similar shapes. This is a retrospective study with 43 out of 107 adult patients that satisfied the inclusion criteria of ICD9 code for SAH, admitted to UCSF Medical Center between 03/2013 and 08/2014, and exclusion criteria of absence of reported EVD clamping trial and ICP signal recordings shorter than 2 hours. By reviewing the clinical notes and pre/post brain imaging results, 25 patients were determined with an intact CSF circulatory system (group A) and 18 patients with an impaired one (group B). Blind to the chart review, the MOCAIP algorithm was applied to the ICP signal recordings to form a series of artifact-free dominant pulses. Then, for each pulse, other pulses with similar mean ICP were identified. Finally, the Euclidean and geodesic inter-pulse distances were calculated in a 4 hours moving window. ANOVA analyses showed a significant difference on the standard deviations of the Euclidean (p<0.001, η2=0.29) and geodesic distance (p=0.001, η2=0.22), between groups A and B. Mann-Whitney U tests were used on the (non-normally distributed) mean distances, showing a significant difference on the Euclidean (p=0.007, r=0.41) and the geodesic distance (p=0.02, r=0.36) between the two groups of patients. The area under the ROC curve was 0.79 and 0.78 for the standard deviations of the Euclidean and geodesic distance, and 0.74 and 0.71 for the mean Euclidean and geodesic distance respectively. Patients with an impaired CSF system showed not only a larger mean but also a larger variability of the inter-pulse distances, indicating frequent changes on the morphology of the pulses. The variability of morphological distances was the best predictor of the effect of EVD clamping trial. This technique may provide a method to rapidly determine if a patient will need surgical placement of an indwelling shunt or can simply have the EVD removed following SAH.

The aim of this study was to provide a non-invasive imaging method to evaluate the physical and mechanical parameters as a novelty method during skin photoaging. In order to evaluate the process of skin damage, 25 mice (C57BL6) were exposed to UVB radiation (0.03mW/cm2 ), 5 times a week for 5weeks. The thickness of the epidermal and dermal layers was measured weekly from the ultrasound images (40MHz). The elastic parameters of the skin were estimated from the processing of the sequential ultrasound images with the motion detection algorithm during the injury generation process. The thickening, Young modulus, and shear modulus of the dermal and epidermal layers during the UVB damage process significantly increased during the 5-week study period (P<.05). In addition, the percentage of changes in the thickness of the epidermal layer (0.22±0.01mm in day 0 to 0.37±0.02mm in day 35) and dermal layer (0.57±0.05mm in day 0 to 0.90±0.08mm in day 35) increased by 68% and 57%, respectively. Furthermore, Young modulus (154.41±8.8kPa) was 11 times more than that of non-irradiated skin (14.90±2.2kPa) and the shear modulus (2.33±0.04kPa) was 2.2 times more than non-irradiated skin (1.06±0.04kPa). With processing the sequential ultrasound images and extracting the thickening, the elasticity of the skin layers can detect skin lesions by UVB radiation.

The main inherent property of medical ultrasound imaging is speckle noise which generally obscures and reduces the diagnostic image resolution and contrast. Consequently, the substantial improvement of ultrasound images is an important prerequisite, whenever ultrasound is used as one of the most utilized diagnostic modalities. The problem of reconstruction of ultrasound images by means of blind deconvolution has long been recognized as a prerequisite in ultrasound imaging processing in the recent decades. Recently, the Total Variation (TV) regularization method has become extremely effective approach for image reconstruction, especially for restoring edges of the blurring image. In this paper, we present a new blind iterative TV deconvolution algorithm for reconstructing ultrasound images from blurry and noisy observations. First, it proposes the initial estimation of the point-spread function (PSF) based on a generalization of two-dimensional homomorphic filtering in cepstrum domain. It is demonstrated that the initial PSF can be effectively estimated by applying a proper smoothing low-pass filtering in cepstrum domain. Second, it introduces a novel blind iterative TV deconvolution which is derived from an alternating minimization algorithm. Fast Fourier Transform (FFT) is used in the pre-iteration computation. The innovative blind deconvolution is based on either concurrent or successive estimation of the PSF function and the image of interest. The iterative scheme is devised to recover the image and simultaneously identify the PSF function. The estimated PSF and restored image will be close to real values in the subsequent iterative deconvolution. Final, tests on phantom and clinical images have proven our novel blind iterative TV deconvolution gives more positive results of higher spatial resolution and better defined tissue structures than other deconvolution methods.

The narrow fields of view obtained from real-time ultrasound transducers, especially with linear array transducers, allow focused evaluation of a specific site but often without any anatomic reference. To allow medical ultrasound imaging to be used in more diverse clinical settings, we have created a new acquisition and display process that allows extended field of view (XFOV) imaging. To produce an XFOV image, extended acoustic slices are obtained by maneuvering the transducer along the body surface or inside. As the images are acquired, they are correlated, aligned, and spliced together into a long composite view, all without the use of a position sensor. This computationally intensive process involves image registration, geometric image transformation, panoramic image construction, and image display. The XFOV process executes in real-time on our programmable ultrasound processing subsystem, the programmable ultrasound image processor, which fits within an existing ultrasound system and supports native ultrasound signal and image processing.

Compared with 2D face recognition systems, 3D facial recognition has been a popular area of study in academia and industry, because of its ability to detect human faces in an unconstrained environment more efficiently. This study presents a new 3D face identification method that uses 3D facial images to compute geodesic distances using 3D landmarks. Feature extraction is computationally more efficient than Euclidean distance using geodesic distance. This study aims to demonstrate how to differentiate identical twins based solely on facial features by developing an efficient algorithm for calculating geodesic paths and distances. We employed the A*, Dijkstra, and fast marching (FM) algorithms to calculate the geodesic distance. Quantitative similarity metrics are obtained and then utilized as inputs in numerous classification methods: random forest (RF), extra tree classifier (ETC), light gradient boosting machine (LGBM), support vector machine (SVM), and bagging classifiers. An expression challenge dataset for 3D twins (3D-TEC), the most challenging data set for 3D face recognition research at Notre Dame University, was used to validate and test the performance of these algorithms. The recall, accuracy, precision, training time, and F1 score were compared for each classifier to determine the best performance. The results from the experiment indicate that, when compared to other algorithms and classification models, the Extra Tree Classifier is the most accurate (by 90%) in predicting face images, whereas A* is the best technique for computing geodesic distance.

Ultrasound (US) image segmentation is a difficult task because of its heavy speckle noise, low quality and blurry boundaries. In this paper, a new neural network based method is proposed for ultrasound images segmentation. A modified self organizing map (SOM) network, named finite impulse response SOM (FIR-SOM), is utilized to segment ultrasound images. A two dimensional (2D) discrete wavelet transform (DWT) is used to build the input feature space of the network. Experimental results show that FIR-SOM discovers the pattern of the input image properly and is robust against noise. Segmentation results of breast ultrasound images (BUS) demonstrate that there is a strong correlation between tumor region selected by a physician and the tumor region segmented by our proposed method.

Automated lesion segmentation in ultrasound (US) images based on deep learning (DL) approaches plays a crucial role in disease diagnosis and treatment. However, the successful implementation of these approaches is conditioned by large-scale and diverse annotated datasets whose obtention is tedious and expertise demanding. Although methods like generative adversarial networks (GANs) can help address sample scarcity, they are often associated with complex training processes and high computational demands, which can limit their practicality and feasibility, especially in resource-constrained scenarios. Therefore, this study is aimed at exploring new solutions to address the challenge of limited annotated samples in automated lesion delineation in US images. Specifically, we propose five distinct mixed sample augmentation strategies and assess their effectiveness using four deep segmentation models for the delineation of two lesion types: breast and thyroid lesions. Extensive experimental analyses indicate that the effectiveness of these augmentation strategies is strongly influenced by both the lesion type and the model architecture. When appropriately selected, these strategies result in substantial performance improvements, with the Dice and Jaccard indices increasing by up to 37.95% and 36.32% for breast lesions and 14.59% and 13.01% for thyroid lesions, respectively. These improvements highlight the potential of the proposed strategies as a reliable solution to address data scarcity in automated lesion segmentation tasks. Furthermore, the study emphasizes the critical importance of carefully selecting data augmentation approaches, offering valuable insights into how their strategic application can significantly enhance the performance of DL models.

In this work, we present a novel technique for ultrasound image processing. The development is carried out trough of the finite impulse response filter (FIR) and the Taylor polynomial approximation. Here, we showed the generic form of the Moving Average Hybrid filter (MAH) to the enhancing ultrasound imaging. Finally, we present any evaluations qualitative and quantitative to the improvement of MAH filter

An acoustic imaging algorithm is proposed herein for transient noise source time reconstruction. Time domain formulations are not well suited for acoustic imaging because of the size of the resulting system to be inversed. Based on the phase coherence principle widely used in ultrasound imaging and image processing, the first step of the algorithm consists in proposing the phase coherence metric used to reject pixels that are unlikely to contribute to the radiated sound field. This translates in a reduction of the domain size and ill-posedness of the problem. In the second step, the inverse problem is solved using the Tikhonov regularization and the generalized cross-validation to extract the vibration field on the imaging domain. Two test cases are considered: a simulated baffled piston and a panel submitted to a mechanical impact in anechoic conditions. The actual vibration field of the panel is measured with an optical technique for reference. In both numerical and experimental cases, the reconstructed vibration field using the proposed approach compares well with their respective reference. The results confirm that transient excitations can be localized and quantified with the proposed approach, in contrast with the classical time-domain beamforming that dramatically overestimates its magnitude.

The technology of needle biopsy and targeted therapy are minimally invasive medical treatments which have many prospects in application. Because of the deformation of biopsy tissue and the bending of flexible needle, it's difficult to accurately control the needle trajectory. For this, we propose an algorithm to determine automatic needle insertion based on ultrasound image processing technology. This method uses image processing algorithms to locate the focal region, then applies snake algorithm (active contour models) and concavity and convexity of boundary to analyze the border traits of tumor region; finally, with the implementation of evaluating relative position of the needle and tumor region, and whether it's suitable to puncture in the focal region around the possible needle puncture point, the prediction of adjusting the angle of the needle puncture is realized. This method provides an effective guide for the realization of automatic needle puncture based on ultrasound image processing.

A new two-stream network based on feature separation and complementation for ultrasound image segmentation

This study aims to enhance the efficiency and accuracy of thyroid nodule segmentation in ultrasound images, ultimately improving nodule detection and diagnosis. For clinical deployment on mobile and embedded devices, DeepLabV3+ strives to achieve a balance between a lightweight architecture and high segmentation accuracy. A comprehensive dataset of ultrasound images was meticulously curated using a high-resolution ultrasound imaging device. Data acquisition adhered to standardized protocols to ensure high-quality imaging. Preprocessing steps, including noise reduction and contrast optimization, were applied to enhance image clarity. Expert radiologists provided ground truth labels through meticulous annotation. To improve segmentation performance, we integrated MobileNetV2 and Depthwise Separable Dilated Convolution into the Atrous Spatial Pyramid Pooling (ASPP) module, incorporating the Pyramid Pooling Module (PPM) and attention mechanisms. To mitigate classification imbalances, we employed Tversky loss functions in the ultrasound image classification process. In semantic image segmentation, DeepLabV3+ achieved an impressive Intersection over Union (IoU) of 94.37%, while utilizing only 12.4 MB of parameters, including weights and biases. This remarkable accuracy demonstrates the effectiveness of our approach. A high IoU value in medical imaging analysis reflects the model's ability to accurately delineate object boundaries. DeepLabV3+ represents a significant advancement in thyroid nodule segmentation, particularly for thyroid cancer screening and diagnosis. The obtained segmentation results suggest promising directions for future research, especially in the early detection of thyroid nodules. Deploying this algorithm on mobile devices offers a practical solution for early diagnosis and is likely to improve patient outcomes.