Mathematics and Machine Learning for Visual Computing in Medicine: Acquisition, Processing, Analysis, Visualization, and Interpretation of Visual Information

As an important branch of artificial intelligence, computer vision plays a huge role in the rapid development of artificial intelligence. From a biological point of view, in the acquisition and processing of information, vision is much more important than hearing, touch, etc., because 70% of the human cerebral cortex is processing visual information. Therefore, advances in computer vision technology are critical to the development of artificial intelligence that is designed to allow machines to think and handle things like humans. The acquisition and processing of visual information has always been the focus of computer vision research, and it is also difficult. The main problem of traditional computer vision technology in the processing of visual information is that the extracted image features are less discriminative, the generalization ability of image features in complex background scenes is insufficient, and the recognition ability on object recognition is poor. In response to these problems, based on the visual neural mechanism, this paper establishes an appropriate computer model for the neuronal cells in the human primary visual cortex, models the recognition response mechanism of the visual ventral system, and performs image feature extraction on the training samples. And object recognition. The results show that compared with the traditional methods, the proposed method effectively improves the discrimination of image features, and the image features extracted under complex background scenes have good generalization ability. On this basis, the training samples can be effectively recognized. The results show that the model based on the visual neural mechanism, the recognition of the edge, orientation and contour of the training sample show the advantages of the biological vision mechanism in object recognition.

The past several decades have witnessed dramatic advances in the technology of image acquisition and display. There have been some parallel but less spectacular advances in understanding human acquisition and analysis of visual information. For the first time, imaging devices are available that allow the operator to determine the brightness relationships between normal and pathologic structures. For the first time too, it is becoming possible for the community of scientists interested in medical imaging to marry knowledge of human observer performance with the capabilities of the modern display devices to permit maximum information extraction from medical images.

Image understanding in the brain or a computer requires segmentation of observed images, i.e., their partition into different semantically-connected parts that each constitute one physical object. This task is fundamental for further processing and analysis of visual information and seems to be accomplished by the brain very easily. Nevertheless it is a very demanding challenge for computer algorithms.

The article shows the importance of solving various production problems using processing, analysis and display of visual information. The existing systems for processing visual information are considered and the insufficiently developed use of automated image processing systems in production systems that solve rather complex and “intelligent” problems of image analysis is noted. The prospects of using digital modeling of object images are shown. The article provides a literary review of works on the development of methods and models for effective processing of visual control information, which showed that there are many unresolved issues. This work is aimed at solving some of them. The relevance of the topic and the objectives of the research are determined. Based on the results of the studies performed, a model of the visual information processing system was developed. From the object under study, a video stream arrives at the input of the model of the visual information processing system, which is first selected as a sequence of frames (frames) of a fixed dimension. Then there is a preliminary processing of the information received from the object, its segmentation, extraction of features and decision making. Here, the traditional number of homogeneous segments can be applied, into which the object is divided using the principle of maximum posterior probability. The result obtained is then used in the adaptation block. The proposed model of the visual information processing system was implemented in the Qt Creator environment in C ++.

1. In squirrel monkeys or macaques under barbiturate anesthesia one or a few electrical pulses applied to the mesencephalic reticular formation or pretectal area alters synaptic transmission through the lateral geniculate nucleus, as tested by magnitude of response in optic radiation to stimulation of optic tract. The effect begins within 5–10 msec after the start of the mesencephalic stimulation and is usually biphasic. Initial inhibition lasts for about 30 msec and is then replaced by facilitation which peaks around 70–120 msec and may endure as much as 1 sec. 2. The major site of the effect is upon the lateral geniculate nucleus, and both magno- and parvocellular elements are involved. Only rarely does the mesencephalic stimulation directly alter excitability of striate cortex. 3. In this respect, as well as in the strong facilitation of photically elicited responses, the effects in primates differ significantly from those in cats. 4. Although the inhibition observed in the lateral geniculate nucleus meets criteria commonly believed to signify a presynaptic origin, several paradoxical facts (see Discussion) must be resolved before this interpretation can be secure. 5. It is hypothesized that the observed effects arise primarily from intrusion into a system normally concerned with the acquisition of visual information by ocular saccades. Saccadic gating of transmission through the lateral geniculate nucleus would distinguish self-generated visual signals from similar, extrinsically initiated signals, and would thus constitute the “corollary discharge”.

Most image and video compression algorithms that have been proposed to improve picture quality relative to compression efficiency have either been designed based on objective criteria such as signal-to-noise-ratio (SNR) or have been evaluated, post-design, against competing methods using an objective sample measure. However, existing quantitative design criteria and numerical measurements of image and video quality both fail to adequately capture those attributes deemed important by the human visual system, except, perhaps, at very low error rates. We present a framework for assessing the quality of and determining the efficiency of foveated and compressed images and video streams. Image foveation is a process of nonuniform sampling that accords with the acquisition of visual information at the human retina. Foveated image/video compression algorithms seek to exploit this reduction of sensed information by nonuniformly reducing the resolution of the visual data. We develop unique algorithms for assessing the quality of foveated image/video data using a model of human visual response. We demonstrate these concepts on foveated, compressed video streams using modified (foveated) versions of H.263 that are standard-compliant. We rind that quality vs. compression is enhanced considerably by the foveation approach.

When primates maintain their gaze directed toward a visual target (visual fixation), their eyes display a combination of miniature fast and slow movements. An involvement of the cerebellum in visual fixation is indicated by the severe gaze instabilities observed in patients suffering from cerebellar lesions. Recent studies in non-human primates have identified a cerebellar structure, the fastigial oculomotor region (FOR), as a major cerebellar output nucleus with projections toward oculomotor regions in the brain stem. Unilateral inactivation of the FOR leads to dysmetric visually guided saccades and to an offset in gaze direction when the animal fixates a visual target. However, the nature of this fixation offset is not fully understood. In the present work, we analyze the inactivation-induced effects on fixation. A novel technique is adopted to describe the generation of saccades when a target is being fixated (fixational saccades). We show that the offset is the result of a combination of impaired saccade accuracy and an altered encoding of the foveal target position. Because they are independent, we propose that these two impairments are mediated by the different projections of the FOR to the brain stem, in particular to the deep superior colliculus and the pontomedullary reticular formation. Our study demonstrates that the oculomotor cerebellum, through the activity in the FOR, regulates both the amplitude of fixational saccades and the position toward which the eyes must be directed, suggesting an involvement in the acquisition of visual information from the fovea.

Based on the criteria set out in Annex 1 of the European Commission (EC) Regulation No.800 / 2008 and official data, about 99% would be in the category of small and medium-sized enterprises (SMEs). In Latvia, the breakdown of economically active SMEs by the Ministry of Economics is: micro enterprises 90%, small enterprises 9%, medium enterprises 1%. The European Commission believes that it is SMEs and entrepreneurship that are key to ensuring economic growth, innovation, job creation and social inclusion in the European Union. However, globalization is one of the ways micro and small businesses can go beyond national territories to expand their business to succeed. The aim of the study is to analyze the types of activity in which micro and small enterprises in Latvia operate for 5 years and to analyze the level of potential readiness of globalization. Quantitative research is based on official business data from several sources Enterprise Register, Ministry of Economics, Re-Identifier of the Register of Enterprises Ltd. “Lursoft”, Central Statistical Bureau. The study will review data on the last 5 years by analyzing data for 2013-2017. The following research methods have been used for the research: quantitative data research approach, analytical study comparing data for five years and graphical method visualization and analysis of visual information.

Brain signal complexity rises with repetition suppression in visual learning

This study explored whether benzodiazepines selectively affect aspects of attention and/or visual information processing, as they do memory. A cued visual-search paradigm was employed, using normal volunteers and a single dose of triazolam. This paradigm provided for a detailed examination of two aspects of visual attention and information processing: 1) controlled versus automatic attention allocation (via central and peripheral cues), and 2) the extent to which processing an item in a non-cued location affects performance (via cue-validity). Triazolam, compared to placebo, significantly increased response time, and Drug Condition interacted with Cue-Validity but not Cue-Type. Based on these data, we argue that triazolam does not affect attention allocation but does affect attentional disengagement and/or attention switching mechanisms.

The article is devoted to a comprehensive study of the application of artificial intelligence in forensic practice, covering its theoretical foundations, historical development and modern methodologies. The interdisciplinary nature of AI and various approaches to its definition are considered, in particular in the context of European Union legislation. The key stages of the development of AI as a scientific discipline are analyzed, from symbolic AI to the era of deep learning, and their impact on forensics. The main emphasis is placed on the classification and detailed description of AI methods used in forensics, including machine learning (supervised, unsupervised, with reinforcement), deep learning (convolutional and recurrent neural networks), natural language processing (NLP) and computer vision. Specific examples of the application of these technologies for the analysis of visual and textual information, biometric identification, crime prediction and war crimes investigation are considered. The article analyzes the international experience of applying AI in forensics, highlighting both successful cases (the use of biometric systems, acoustic systems for detecting gunshots, analytical tools for searching for criminals), and problematic situations that led to illegal arrests and erroneous convictions due to the bias of algorithms or insufficient accuracy of technologies. Based on international experience, recommendations are formulated for the implementation of AI in forensics in Ukraine. The need for thorough testing, the development of a clear regulatory framework, ensuring the transparency of algorithms, independent expert assessment and human supervision is emphasized. Particular attention is paid to the ethical aspects of the use of AI and its potential impact on human rights. A multifaceted approach to reducing risks and optimizing the use of AI is proposed, including the formation of an ethical and legal framework, ensuring the transparency and accountability of algorithms, assessing and eliminating bias, improving the quality of training data, and establishing interdisciplinary cooperation. The importance of training qualified specialists to work with AI systems is emphasized.

Mobile and Web-of-Things (WoT) devices at the network edge account for more than half of the world's web traffic, making a great data source for various machine learning (ML) applications, particularly federated learning (FL) which offers a promising solution to privacy-preserving ML feeding on these data. FL allows edge mobile and WoT devices to train a shared global ML model under the orchestration of a central parameter server. In the real world, due to resource heterogeneity, these edge devices often train different versions of models (e.g., VGG-16 and VGG-19) or different ML models (e.g., VGG and ResNet) for the same ML task (e.g., computer vision and speech recognition). Existing FL schemes have assumed that participating edge devices share a common model architecture, and thus cannot facilitate FL across edge devices with heterogeneous ML model architectures. We explored this architecture heterogeneity challenge and found that FL can and should accommodate these edge devices to improve model accuracy and accelerate model training. This paper presents our findings and FlexiFed, a novel scheme for FL across edge devices with heterogeneous model architectures, and three model aggregation strategies for accommodating architecture heterogeneity under FlexiFed. Experiments with four widely-used ML models on four public datasets demonstrate 1) the usefulness of FlexiFed; and 2) that compared with the state-of-the-art FL scheme, FlexiFed improves model accuracy by 2.6%-9.7% and accelerates model convergence by 1.24 × -4.04 ×.

The unmanned aerial vehicles (UAVs) have been extensively used in providing intelligence such as target tracking. In our field experiments, a pre-trained deep neural network (DNN) is deployed on the UAV to identify a target from the captured video frames and enable UAV to keep tracking. However, tracking in real time by the DNN requires a lot of computational resources. This motivates us to consider offloading this type of machine learning (ML) tasks to a mobile edge computing (MEC) server. Specifically, we propose a novel hierarchical ML tasks distribution framework for the UAV tracking system, where the UAV is embedded with lower layers of the pre-trained convolutional neural network (CNN) model due to its limited computing capability, while the MEC server with rich computing resources will handle the higher layers of the CNN model. An optimization problem is formulated to minimize the CNN inference delay while taking into account the communications delay, computing time, and ML error. Insights are provided to understand the tradeoff between communications and ML computing in offloading decisions. Numerical results demonstrate the effectiveness of the proposed ML tasks distribution framework with the optimized offloading strategy.

A birth cohort observational study to assess the association of prenatal biomarkers of organophosphates with visual attention, recognition memory and information processing among Thai infants

The relevance is determined by the key problem of modern intelligent systems — the limited effectiveness of existing computer vision models for predicting dynamic processes in changing conditions. The need to develop high-tech technologies capable of analyzing multidimensional time series and ensuring high reliability of forecasts is a fundamental challenge in the field of artificial intelligence. The aim of the research is to develop methodological foundations for creating predictive computer vision models based on hybrid deep learning architectures for time series analysis of multimodal data. Methods. Hybrid deep learning architectures, including combinations of convolutional (CNN) and recurrent neural networks (LSTM), as well as ConvLSTM models, have been investigated. A mathematical apparatus has been developed to formalize learning processes, including specialized loss functions and optimization algorithms adapted for forecasting tasks based on time series of visual data. Results. It has been experimentally shown that the CNN-LSTM hybrid model predicts the dynamics of the development of objects 8 time steps ahead with an average absolute error of 0.069 and a coefficient of determination of 0.9127. A statistically significant superiority of this architecture over isolated convolutional or recurrent networks has been established. The effectiveness of the LSTM architecture for predicting anomalous events based on multidimensional data with an accuracy of 0.92 has been proven. The prospects of using multimodal and graph neural network models to increase the reliability of forecasts by taking into account spatial and temporal relationships are revealed. Practical significance. The results of the work form the basis for the creation of high-tech information and analytical systems of predictive analytics, applicable in various subject areas where the analysis of visual information with a timestamp is required. Discussion. The developed approaches demonstrate the potential for creating intelligent anomaly detection and decision support systems. The development of methods that are resistant to a limited amount of training data and the creation of approaches to quantify forecast uncertainty in complex dynamic environments are identified as promising areas.