Search

Rationale and Objectives: This study investigated the use of deep learning-generated virtual positron emission tomography (PET)-like gated single-photon emission tomography (SPECT VP ) for assessing myocardial strain, overcoming limitations of conventional SPECT., Materials and Methods: SPECT-to-PET translation models for short-axis, horizontal, and vertical long-axis planes were trained using image pairs from the same patients in stress (720 image pairs from 18 patients) and resting states (920 image pairs from 23 patients). Patients without ejection-fraction changes during SPECT and PET were selected for training. We independently analyzed circumferential strains from short-axis-gated SPECT, PET, and model-generated SPECT VP images using a feature-tracking algorithm. Longitudinal strains were similarly measured from horizontal and vertical long-axis images. Intraclass correlation coefficients (ICCs) were calculated with two-way random single-measure SPECT and SPECT VP (PET). ICCs (95% confidence intervals) were defined as excellent (≥0.75), good (0.60-0.74), moderate (0.40-0.59), or poor (≤0.39)., Results: Moderate ICCs were observed for SPECT-derived stressed circumferential strains (0.56 [0.41-0.69]). Excellent ICCs were observed for SPECT VP -derived stressed circumferential strains (0.78 [0.68-0.85]). Excellent ICCs of stressed longitudinal strains from horizontal and vertical long axes, derived from SPECT and SPECT VP , were observed (0.83 [0.73-0.90], 0.91 [0.85-0.94])., Conclusion: Deep-learning SPECT-to-PET transformation improves circumferential strain measurement accuracy using standard-gated SPECT. Furthermore, the possibility of applying longitudinal strain measurements via both PET and SPECT VP was demonstrated. This study provides preliminary evidence that SPECT VP obtained from standard-gated SPECT with postprocessing potentially adds clinical value through PET-equivalent myocardial strain analysis without increasing the patient burden., Competing Interests: Declaration of Competing Interest The authors declare no competing interests., (Copyright © 2024 The Association of University Radiologists. Published by Elsevier Inc. All rights reserved.)

Purpose: Synthetic datasets hold the potential to offer cost-effective alternatives to clinical data, ensuring privacy protections and potentially addressing biases in clinical data. We present a method leveraging such datasets to train a machine learning algorithm applied as part of a computer-aided detection (CADe) system., Approach: Our proposed approach utilizes clinically acquired computed tomography (CT) scans of a physical anthropomorphic phantom into which manufactured lesions were inserted to train a machine learning algorithm. We treated the training database obtained from the anthropomorphic phantom as a simplified representation of clinical data and increased the variability in this dataset using a set of randomized and parameterized augmentations. Furthermore, to mitigate the inherent differences between phantom and clinical datasets, we investigated adding unlabeled clinical data into the training pipeline., Results: We apply our proposed method to the false positive reduction stage of a lung nodule CADe system in CT scans, in which regions of interest containing potential lesions are classified as nodule or non-nodule regions. Experimental results demonstrate the effectiveness of the proposed method; the system trained on labeled data from physical phantom scans and unlabeled clinical data achieves a sensitivity of 90% at eight false positives per scan. Furthermore, the experimental results demonstrate the benefit of the physical phantom in which the performance in terms of competitive performance metric increased by 6% when a training set consisting of 50 clinical CT scans was enlarged by the scans obtained from the physical phantom., Conclusions: The scalability of synthetic datasets can lead to improved CADe performance, particularly in scenarios in which the size of the labeled clinical data is limited or subject to inherent bias. Our proposed approach demonstrates an effective utilization of synthetic datasets for training machine learning algorithms., (© 2024 The Authors.)

Convolutional neural networks have achieved great success in computer vision, but incorrect predictions would be output when applying intended perturbations on original input. These human-indistinguishable replicas are called adversarial examples, which on this feature can be used to evaluate network robustness and security. White-box attack success rate is considerable, when already knowing network structure and parameters. But in a black-box attack, the adversarial examples success rate is relatively low and the transferability remains to be improved. This article refers to model augmentation which is derived from data augmentation in training generalizable neural networks, and proposes resizing invariance method. The proposed method introduces improved resizing transformation to achieve model augmentation. In addition, ensemble models are used to generate more transferable adversarial examples. Extensive experiments verify the better performance of this method in comparison to other baseline methods including the original model augmentation method, and the black-box attack success rate is improved on both the normal models and defense models., Competing Interests: The authors declare that they have no competing interests., (© 2023 Li et al.)

Objective: To propose a framework that combines sinogram interpolation with unsupervised image-to-image translation (UNIT) network to correct metal artifacts in CT images., Methods: The initially corrected CT image and the prior image without artifacts, which were considered as different elements in two different domains, were input into the image transformation network to obtain the corrected image. Verification experiments were carried out to assess the effectiveness of the proposed method using the simulation data, and PSNR and SSIM were calculated for quantitative evaluation of the performance of the method., Results: The experiment using the simulation data showed that the proposed method achieved better results for improving image quality as compared with other methods, and the corrected images preserved more details and structures. Compared with ADN algorithm, the proposed algorithm improved the PSNR and SSIM by 2.4449 and 0.0023 when the metal was small, by 5.9942 and 8.8388 for images with large metals, and by 8.8388 and 0.0130 when both small and large metals were present, respectively., Conclusion: The proposed method for metal artifact correction can effectively remove metal artifacts, improve image quality, and preserve more details and structures on CT images.

The scarcity of training annotation is one of the major challenges for the application of deep learning technology in medical image analysis. Recently, self-supervised learning provides a powerful solution to alleviate this challenge by extracting useful features from a large number of unlabeled training data. In this article, we propose a simple and effective self-supervised learning method for leukocyte classification by identifying the different transformations of leukocyte images, without requiring a large batch of negative sampling or specialized architectures. Specifically, a convolutional neural network backbone takes different transformations of leukocyte image as input for feature extraction. Then, a pretext task of self-supervised transformation recognition on the extracted feature is conducted by a classifier, which helps the backbone learn useful representations that generalize well across different leukocyte types and datasets. In the experiment, we systematically study the effect of different transformation compositions on useful leukocyte feature extraction. Compared with five typical baselines of self-supervised image classification, experimental results demonstrate that our method performs better in different evaluation protocols including linear evaluation, domain transfer, and finetuning, which proves the effectiveness of the proposed method., (© 2022 Wiley-VCH GmbH.)

Anomalies are a set of samples that do not follow the normal behavior of the majority of data. In an industrial dataset, anomalies appear in a very small number of samples. Currently, deep learning-based models have achieved important advances in image anomaly detection. However, with general models, real-world application data consisting of non-ideal images, also known as poison images, become a challenge. When the work environment is not conducive to consistently acquiring a good or ideal sample, an additional adaptive learning model is needed. In this work, we design a potential methodology to tackle poison or non-ideal images that commonly appear in industrial production lines by enhancing the existing training data. We propose Hierarchical Image Transformation and Multi-level Features (HIT-MiLF) modules for an anomaly detection network to adapt to perturbances from novelties in testing images. This approach provides a hierarchical process for image transformation during pre-processing and explores the most efficient layer of extracted features from a CNN backbone. The model generates new transformations of training samples that simulate the non-ideal condition and learn the normality in high-dimensional features before applying a Gaussian mixture model to detect the anomalies from new data that it has never seen before. Our experimental results show that hierarchical transformation and multi-level feature exploration improve the baseline performance on industrial metal datasets.

Visual properties that primarily attract bottom-up attention are collectively referred to as saliency. In this study, to understand the neural activity involved in top-down and bottom-up visual attention, we aim to prepare pairs of natural and unnatural images with common saliency. For this purpose, we propose an image transformation method based on deep neural networks that can generate new images while maintaining the consistent feature map, in particular the saliency map. This is an ill-posed problem because the transformation from an image to its corresponding feature map could be many-to-one, and in our particular case, the various images would share the same saliency map. Although stochastic image generation has the potential to solve such ill-posed problems, the most existing methods focus on adding diversity of the overall style/touch information while maintaining the naturalness of the generated images. To this end, we developed a new image transformation method that incorporates higher-dimensional latent variables so that the generated images appear unnatural with less context information but retain a high diversity of local image structures. Although such high-dimensional latent spaces are prone to collapse, we proposed a new regularization based on Kullback-Leibler divergence to avoid collapsing the latent distribution. We also conducted human experiments using our newly prepared natural and corresponding unnatural images to measure overt eye movements and functional magnetic resonance imaging, and found that those images induced distinctive neural activities related to top-down and bottom-up attentional processing., Competing Interests: Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper., (Copyright © 2022 The Author(s). Published by Elsevier Ltd.. All rights reserved.)

The synthesis between face sketches and face photos has important application values in law enforcement and digital entertainment. In cases of a lack of paired sketch-photo data, this paper proposes an unsupervised model to solve the problems of missing key facial details and a lack of realism in the synthesized images of existing methods. The model is built on the CycleGAN architecture. To retain more semantic information in the target domain, a multi-scale feature extraction module is inserted before the generator. In addition, the convolutional block attention module is introduced into the generator to enhance the ability of the model to extract important feature information. Via CBAM, the model improves the quality of the converted image and reduces the artifacts caused by image background interference. Next, in order to preserve more identity information in the generated photo, this paper constructs the multi-level cycle consistency loss function. Qualitative experiments on CUFS and CUFSF public datasets show that the facial details and edge structures synthesized by our model are clearer and more realistic. Meanwhile the performance indexes of structural similarity and peak signal-to-noise ratio in quantitative experiments are also significantly improved compared with other methods.

Machine Learning (ML) methods have become state of the art in radar signal processing, particularly for classification tasks (e.g., of different human activities). Radar classification can be tedious to implement, though, due to the limited size and diversity of the source dataset, i.e., the data measured once for initial training of the Machine Learning algorithms. In this work, we introduce the algorithm Radar Activity Classification with Perceptual Image Transformation (RACPIT), which increases the accuracy of human activity classification while lowering the dependency on limited source data. In doing so, we focus on the augmentation of the dataset by synthetic data. We use a human radar reflection model based on the captured motion of the test subjects performing activities in the source dataset, which we recorded with a video camera. As the synthetic data generated by this model still deviates too much from the original radar data, we implement an image transformation network to bring real data close to their synthetic counterpart. We leverage these artificially generated data to train a Convolutional Neural Network for activity classification. We found that by using our approach, the classification accuracy could be increased by up to 20%, without the need of collecting more real data.

Transmission electron microscope (TEM) specimen preparation by focused ion beam (FIB) milling requires delicate polishing of a thin window of material during the final stages of the procedure. Over or underpolishing is common and requires extra microscope resources to correct. Despite some methods for lamella thickness measurement being available, the majority of users judge the final polishing step subjectively from scanning electron microscope (SEM) images acquired between milling steps. Here we demonstrate successful thickness determination of thin silicon lamellae using calibrated secondary electron detectors in a FIB-SEM dual-beam chamber. Unlike previous thickness measurement methods it does not require long acquisition times, the use of in-chamber scanning transmission electron microscope (STEM) or energy dispersive x-ray spectroscopy detectors. The calibration aligns a SEM image to an electron energy loss spectroscopy (EELS) map of lamella thickness acquired in a TEM. This calibration reveals the greyscale-thickness dependence of two secondary electron SEM detectors: the through-lens detector (TLD) and the in-chamber electron detector (ICE). It was found that lamella thickness estimation for TLD images is accurate for areas thinner than 0.4 t/λ, whilst ICE images are most accurate for areas thicker than 0.5 t/λ up to 1.1 t/λ. The procedure presented here allows objective lamella thickness determination during the final stages of FIB specimen preparation using conventional imaging modes for common secondary electron detectors. LAY DESCRIPTION: Successful analysis of a material in a transmission electron microscope requires very thin windows of the material to be fabricated. Despite the quality of this analysis relying heavily on the thickness of the window, measuring thickness during window fabrication is not common practice. The authors show that it is possible to measure the thickness of the window directly in a focused-ion-beam chamber with a scanning electron microscope without altering the fabrication procedure, and using electron detectors common to most microscopes., (© 2019 Royal Microscopical Society.)

Objective: Mental imagery has featured prominently in several disorders. The recurrent and distressing imagery reported by individuals with social phobia is often that of a previously experienced event and such imagery is proposed to contribute to the maintenance of the disorder in contemporary cognitive models. The aims of the present study were to investigate which imagery processes (generation, maintenance, inspection and transformation) are most relevant to differentiate the performances of socially anxious individuals from those who are non-socially anxious, and that play a pivotal role in social phobia., Method: Sixty-two undergraduate volunteer students were recruited from an Italian university, age range 18-23 (M 19.34, SD 1.27), 23 males and 39 females. We used a standardized battery, the Mental Imagery Test (MIT), measuring mental imagery skills involving generation and manipulation of different categories of images to examine imagery processes in individuals with social anxiety and non-anxious controls., Results: Results shown that socially anxious students were impaired in MIT total score, and particularly in the imagined paths tasks and clock test, significantly more than non-anxious individuals. In overall group, total score was negatively associated with social anxiety, and the imagine paths test was the best predictor of social anxiety., Conclusions: The current findings suggest that social anxiety is differently influenced by specific tasks of imagery based on the type of component involved in imaginative processes., Competing Interests: Declaration of interest: none, (© 2019 Giovanni Fioriti Editore s.r.l.)

Image transformation between multiple domains has become a challenging problem in deep generative networks. This is because, in real-world applications, finding paired images in different domains is an expensive and impractical task. This paper proposes a new model named joint moment-matching autoencoders(JMA). This model learns to perform cross-domain transformation over multiple domains based on perceptual loss and maximum mean discrepancy criteria, in the absence of any paired images between the domains. Our results show that the proposed JMA framework successfully learns to transform images between domains without any paired data. We demonstrate that our model has good performance in the generative context as well as in the domain transformation tasks with better computational efficiency than conventional methods., (Copyright © 2018 Elsevier Ltd. All rights reserved.)

Coupled Generative Adversarial Network (CoGAN) was recently introduced in order to model a joint distribution of a multi modal dataset. The CoGAN model lacks the capability to handle noisy data as well as it is computationally expensive and inefficient for practical applications such as cross-domain image transformation. In this paper, we propose a new method, named the Coupled Generative Adversarial Stacked Auto-encoder (CoGASA), to directly transfer data from one domain to another domain with robustness to noise in the input data as well to as reduce the computation time. We evaluate the proposed model using MNIST and the Large-scale CelebFaces Attributes (CelebA) datasets, and the results demonstrate a highly competitive performance. Our proposed models can easily transfer images into the target domain with minimal effort., (Copyright © 2018 Elsevier Ltd. All rights reserved.)

Fingerprint evidence can benefit in image quality if transformed using digital image processing techniques. This is especially true when considering prints that cannot be easily lifted (such as those deposited on porous paper substrates), or when the mechanism of lifting does not effectively reduce background interferences. In these instances, frequency filtering is one type of mathematical transformation that can serve to increase image clarity and the ability to extract minutiae relevant to pairwise comparisons. To quantify the impact of frequency filtering on image quality, high quality and low quality (noisy) prints were collected. The high quality prints served as exemplars that were compared to the low quality prints both pre- and post-filtering. The resulting pairwise match scores indicate that: (1) frequency filtering has a low probability of creating false positive associations, (2) 90% of the post-filtered images result in a normalized gain in match score, (3) frequency filtering doubled the probability of obtaining match scores greater than 30% (for the automated algorithm employed in this study), and (4) filtering can double the probability of obtaining 10 or more matching minutiae when comparing same source prints. Overall, the research indicates a reasonable and quantifiable payoff in increased clarity, matching minutiae and pairwise similarity for post-filtered images when compared to known-match exemplars., (Copyright © 2014 Elsevier Ireland Ltd. All rights reserved.)