Edge Intensity Research Articles

AS2LS: Adaptive Anatomical Structure-based Two-layer Level Set Framework for Medical Image Segmentation.

Medical images often exhibit intricate structures, inhomogeneous intensity, significant noise and blurred edges, presenting challenges for medical image segmentation. Several segmentation algorithms grounded in mathematics, computer science, and medical domains have been proposed to address this matter; nevertheless, there is still considerable scope for improvement. This paper proposes a novel adaptive anatomical structure-based two-layer level set framework (AS2LS) for segmenting organs with concentric structures, such as the left ventricle and the fundus. By adaptive fitting region and edge intensity information, the AS2LS achieves high accuracy in segmenting complex medical images characterized by inhomogeneous intensity, blurred boundaries and interference from surrounding organs. Moreover, we introduce a novel two-layer level set representation based on anatomical structures, coupled with a two-stage level set evolution algorithm. Experimental results demonstrate the superior accuracy of AS2LS in comparison to representative level set methods and deep learning methods.

Operando lateral state-of-charge inhomogeneity mapping via wavelength-resolved neutron imaging

The continuous advancement in battery technology requires the development of robust and more precise diagnostic tools for the characterization of commercial-scale battery cells. In this work, operando Bragg edge imaging (BEI), based on a time-of-flight approach, provides spatially resolved state-of-charge (SoC) mapping of a commercial-size pouch cell. The changes in the position and intensity of Bragg edges are used to track the Li intercalation into the graphite, thereby identifying the spatial distribution of various lithiation stages and the corresponding SoC. The operando wavelength-resolved BEI assessment revealed an inhomogeneous Li+ intercalation, dependent on the distance to the electrical tabs, effectively identifying features that contribute to the overall loss of performance and the accelerated degradation of battery materials. The presented results demonstrate the potential of advanced wavelength-resolved neutron imaging to be used as a diagnostic tool for commercial battery cells in a configuration without special casings or isotopes for enhancement of imaging contrast. It thus highlights its applicability for operando characterization of future batteries for portable and automotive applications.

Contrast enhancement of digital images using dragonfly algorithm

Contrast enhancement aims to amplify the visual quality of images by modifying the contrast level because digital images may get distorted by casual acquisition. The article deals with contrast enhancement as an optimization problem and uses the Dragonfly Algorithm (DA) to find the optimal grey-level intensity values. The DA for contrast enhancement uses five control parameters (entropy, number of edges, total intensities of edges, the variance of the probability of occurrence of each grey value, and the number of grey levels) to generate an objective function. An ablation study is also performed to understand how different controlling parameter combinations contribute to determining the optimal solution. The proposed approach considers 24 grey-scale images from the Kodak dataset and metrics as Peak Signal-to-Noise Ratio (PSNR), Visual Information Fidelity (VIF), and Structural Similarity Index Measure (SSIM) to verify the output's performance. The PSNR, VIF, and SSIM values in the experiments are 30.87, 0.7451, and 0.9523, respectively. The experimental observations reveal that the proposed DA-based image contrast enhancement produces high-quality images from its low-contrast counterparts. Comparisons with state-of-art methods ensure the superiority of the proposed algorithm. The Python implementation of the proposed approach is available in this Github repository.

Tunable topological edge state in plasma photonic crystals

In this study, we found a kind of edge state located at the interface between plasma photonic crystals (PPCs) and traditional photonic crystals, which depends on the property of the photonic band gap rather than the surface defect. Simulation and theoretical analysis show that by adjusting the plasma density, we can change the topological characteristics of the photonic band gap of PPCs. This makes it different from the photonic band gap of traditional PCs, and thus excites or closes the topological edge states. We further discussed the influence of plasma parameters on edge state characteristics, and the results showed that as the plasma density increased, the first photonic band gap (PBG) of the PPCs closed and then reopened, resulting in band inversion and a change in the PBG properties of the PPCs. We can control the generation of edge states through plasma and adjust the frequency and strength of the edge states. After the appearance of edge states, as the plasma density further increases, the first PBG of the PPCs will shift towards high frequencies and deepen. The frequency of edge states will shift towards higher frequencies, and their strength will also increase. We increased the first PBG depth of the PPCs by increasing the number of arrays and found that when the number of the PPCs arrays increased, only the intensity of the edge states would increase while the frequency remained unchanged. Therefore, flexible adjustment of edge state frequency and intensity can be achieved through plasma density and array quantity parameters. Our study demonstrates the properties of topological edge states in plasma photonic crystals, which we believe can provide some guidance for applications based on edge states.

An underwater image enhancement method based on multi-scale layer decomposition and fusion

High-quality underwater images can intuitively reflect the most realistic underwater conditions, guiding for underwater environmental monitoring and resource exploration strongly. But when factors like light absorption affect underwater optical imaging, it has been found that poor visibility and blurred texture details occur in acquired images, posing challenges for the identification and detection of underwater targets. To obtain natural images, an enhancement algorithm is proposed based on multi-scale layer decomposition and fusion. The algorithm employs different strategies to recover image attenuation information from both local and global perspectives, generating two complementary preprocessed fusion inputs. For fusion input 1, operations are conducted in the RGB color space. Initially, the mean proportion of each color channel is used to identify the attenuated color channel. Then, a local compensation strategy is adaptively applied to restore the pixel intensity of the attenuated color channel. Finally, a statistical color correction method is used to eliminate color cast in the image. Fusion input 2 involves two processing stages. In the Lab color space, the algorithm uses the grayscale information to reduce the deviation in the mean values of channels a and b globally. The local mean information of the component L enhances detail textures. In the RGB color space, linear stretching is applied to correct color deviations. To fuse the structural features of two complementary preprocessed inputs and avoid interference between signals from different layers, the color channels of fused input image are first decomposed into muti-scale structural layers based on structural priors. Then, the image enhancement is achieved through layer-by-layer fusion of the corresponding color channels of the two inputs. By testing and analyzing with the, it was found that the proposed method can improve the clarity of attenuated images in various underwater scenarios of UIEBD and RUIE datasets effectively, enhancing image detail and texture richness, increases contrast, and achieving natural and comfortable visual quality. Compared with the quantitative metrics of 14 other algorithms, the proposed algorithm shows an average score improvement of 10.14, 90.48, and 2.06, respectively, in metrics AG (average gradient), EI (edge intensity), and NIQE (natural image quality evaluator). In the RUIE dataset, it shows an average score improvement of 10.21, 94.76, and 1.86, respectively.

Evaluation of hair surface structure and morphology of patients with lichen planopilaris (LPP) by atomic force microscopy (AFM).

Lichen planopilaris (LPP) is a chronic lymphocytic skin disease manifested by progressive scarring alopecia. The diagnosis of LPP is made based on histopathological examination, although it is not always definite. The current study evaluates the effectiveness of non-invasive atomic force microscopy (AFM) hair examination in detecting morphological differences between healthy and diseased hair. Here, three to five hairs from lesional skin of 10 LPP patients were collected and examined at nine locations using AFM. At least four images were taken at each of the nine sites. Metric measurements were taken and metric (length, width, and scale step height) and morphological features (striated and smooth surface of scales, the presence of endocuticle and cortex, shape of scales edges, scratches, pitting, cracks, globules, and wavy edge) were compared with hair from healthy controls. In addition, areas on diseased hair where the process of pathological, unnatural delamination of the hair fiber occurs are described. There was a statistically significant difference in the number of scratches in the initial sections of the LPP hair, in the intensity of wavy edges along the entire length of the tested hair, and in the number of scales with pitting in the middle section of the hair. In addition, a statistically significant higher number of scales with striated surface was found in LPP group starting at 3.5cm from the root continuing towards the free end of the hair. Other morphological changes such as presence of cortex, globules, oval indentations, and rod-like macrofibrillar elements were also assessed, however, detailed results are not presented, as the differences shown in the number of these morphological changes were not significantly different. This publication outlines the differences between virgin, healthy Caucasian hair, and the hair of LPP patients. The results of this study can be used for further research and work related to LPP. This is the first attempt to characterize the hair of LPP patients using AFM.

The association between online learning, perceived parental relationship, anxiety, and depression symptoms among secondary school students: insight from symptom network and cross-lagged panel network approach

IntroductionAnxiety and depression often co-occur in adolescents, with factors from family and school playing a significant role in the comorbidity. However, network analysis has not examined and clarified the detailed bridge and central symptoms of this comorbidity caused by online learning and perceived parental relationships across different COVID-19 times.MethodsOver four months, 2,356 secondary school students completed the Patient Health Questionnaire-9 and Generalized Anxiety Disorder Scale-7. Participants were divided into harmonious and disharmonious groups based on their answers to a question about parental conflicts.ResultsThe results indicated that adolescents perceiving more parental conflicts showed a denser comorbidity network after four months of online learning. Significant bridge symptoms decreased from three to two across two waves in the harmonious group, while in the disharmonious group, they increased from two to three. The number of central symptoms increased from one in wave 1 to three in wave 2 for the harmonious group, while four in wave 1 decreased to two in wave 2 for the disharmonious group. Furthermore, the CLPN analysis revealed that the strongest positive cross-lagged edge intensity between symptoms was anhedonia—energy in the harmonious group, with anhedonia being the most trigger symptom. In contrast, for the disharmonious group, guilt—suicide and trouble relaxing—excessive worry were the strongest cross-lagged edge, and trouble relaxing was the most trigger symptom.ConclusionThese findings may have implications for interventions designed to promote adolescent mental health in the context of online learning and parental conflicts.

Impact of oxygen plasma power on the performance of Ga2O3 passivated GaN ultraviolet photodetectors

The role of oxygen plasma power on the extent of surface passivation of Gallium Nitride (GaN) epitaxial layers and the photo-response of Au/Ni/GaN Schottky detectors is investigated. The surface morphology is found to be preserved in oxide passivated GaN at intermediate plasma power which significantly deteriorates at higher power. The photoluminescence spectrum shows a fall in the intensity of band edge and defects related yellow luminescence at higher power indicating the generation of non-radiative recombination centres. The chemical and electronic structures of the passivated layer investigated by x-ray photoelectron spectroscopy revealed the formation of stoichiometric Ga2O3 up to intermediate plasma power and partial decomposition of Ga2O3 along with formation of gallium sub-oxides at higher power. This leads to an enhancement in the responsivity of the photodetector after passivation at an intermediate plasma power followed by a considerable reduction at higher power. From the electrical transport measurement across the Au/Ni/GaN Schottky diode, it is observed that the diode manifests more Schottky (Ohmic) character after oxygen plasma treatment at intermediate (higher) power. The present investigations provide a clear guideline on the selection of oxygen plasma power for surface passivation on GaN, which is highly beneficial in optimizing the performance of GaN-based optoelectronic and power devices.

Photoacoustic Image Analysis of Dental Tissue Using Two Wavelengths: A Comparative Study

This work compares photoacoustic images of a tooth by analyzing the signals generated with wavelengths 532 and 355 nm. This comparison addresses the differences in the optical properties of dental tissue for these wavelengths that affect the resulting photoacoustic images. A pulsed Nd:YAG laser was used to illuminate a complete extracted tooth sample, and 2D photoacoustic images (PAIs) were reconstructed using the single-sensor scanning synthetic aperture focusing technique (SSC-SAFT), which is a suitable method for our experimental system with forward detection mode. Signal comparison was conducted using sinogram, signal-to-noise ratio (SNR), root mean square (RMS), arrival time, maximum amplitude, and fast Fourier transform (FFT). PAI comparison utilized intensity profile, edge correlation, and image composition tools. The signal analysis revealed that at 532 nm, the signals exhibited longer decay time and a wider distribution of vibration frequencies due to higher laser pulse energy and greater optical penetration depth. Conversely, at 355 nm, the signals had shorter decay times and a lower frequency distribution, which was attributed to lower energy but improved optical absorption, resulting in reconstructed images with better sharpness and contour definition. This study contributes to the advancement of photoacoustic imaging technology in dentistry by providing insights that could optimize signal generation and image reconstruction for dental tissue.

UCP2-ACGAN: An adaptive condition GAN guided by U-shaped context perceptual processor for PET/CT images fusion

In order to better highlight the lesion area information and edge detailed information of the fused images, an adaptive condition Generative Adversarial Network (GAN) guided by U-shaped context perceptual processor (UCP2-ACGAN) for Positron Emission Tomography (PET) and Computed Tomography (CT) images fusion is proposed in this paper. Firstly, the network architecture of single generator and dual relative discriminators with weights sharing is used in UCP2-ACGAN. Secondly, a multi-granularity adaptive feature extraction module from coarse to fine (MGAFE) is proposed in generator, it is used to extract the valid information contained in the source images. Thirdly, a dual branch feature interactive fusion module guided by U-shaped context perceptual processor (UCP2-DBFIF) is proposed in generator. The context perceptual feature maps are obtained by U-shaped context perceptual processor (UCP2) and multiply them with the feature extraction results as the guiding conditions in the fusion process. In addition, in order to supplement the effective information lost during the feature fusion process, this paper converges the feature information of different granularity during the feature extraction stage. Finally, compared to the best of the 6 comparison methods, the results of comparison experiments show that UCP2-ACGAN is effective on several evaluation metrics values. For example, on the evaluation metrics of Average Gradient (AG), Edge Intensity (EI), Spatial Frequency (SF), Visual Information Fidelity (VIF) and Mutual Information (MI), the metrics values of CT mediastinal window images and PET fused images average increased by 30.74%, 29.35%, 40.24%, 35.47% and 26.18%.

The Influence of nc-AlCrTiN/α-BN Coatings on Increasing the Durability of WC/Co Cutting Inserts in the Inconel Alloy Machining Process.

Anti-wear coatings obtained through PVD methods may significantly increase the durability of cutting tools by impacting their wear mechanisms. This study presents and discusses the results of studies on the impact of the thermal conductivity of PVD coatings on the intensity of the built-up edge (BUE) and built-up layer (BUL) formation in Inconel 600 alloy machining processes. The authors determine the microstructure, phase structure, mechanical properties (hardness, Young's modulus, and adhesion), and thermal conductivity of different PVD coatings selected for the purpose of the study and varying in terms of conductivity-i.e., AlCrTiN and AlCrTiN/BN. Machining processes were carried out under controlled conditions using VBGT160404-M3 cutting inserts with AlCrTiN and AlCrTiN/BN coatings deposited on their surface. The authors prove that the adjustment of the thermal conductivity of PVD coatings to the thermal conductivity of the tool and machined materials can help change the direction of heat flow to cool the cutting zone more effectively. The study results presented in this article show that the deposition of the AlCrTiN/BN coating reduces the friction wear on the tool flank by over 70% and lowers the intensity of BUE and BUL formation processes on the face by 10%, compared to the AlCrTiN coating.

Novel fusion strategy for image fusion using rescue hunt optimization-based modified guidance model

ABSTRACT A new method for image fusion introduces an inventive strategy for amalgamating data from multiple images, resulting in the creation of a single, improved output image. This approach aims to address the limitations and challenges associated with conventional fusion techniques, paving the way for improved results in various applications. In this research, a novel approach for image fusion is presented, featuring a rescue hunt optimisation-based modified guidance model (RHO-based MG model). The methodology leverages the non-subsampled contourlet transform (NSCT) and non-subsampled shear let transform (NSST) to construct the fusion transform using two input images, typically infrared and visual images. By hybridising high-frequency (HF) and low-frequency (LF) bands from both types of images, the fusion model generates the final HF and LF bands. A distinctive modified guidance strategy is employed in the development of these bands. The proposed approach utilises a rescue hunt algorithm developed through the combination of search and rescue optimisation (SARO) and grey wolf optimisation (GWO) behaviours. In image fusion, optimisation fine-tunes VGG-19 and RESNET 18 models to improve their ability to combine multiple images effectively. This process involves adjusting the models’ internal parameters using optimisation techniques. By analysing the features in different input images, these models learn to extract meaningful information and create a fused image that retains the important details from each source. This strategy is further enhanced by integrating the VGG-19 and RESNET 18 models. The fused image is composed of combined HF and LF bands, with the final result obtained through an inverse hybrid transform. Experimental results, conducted on a dataset of 25 images, demonstrate the effectiveness of the approach with metrics average gradient (AG), edge intensity (EI), PSNR, RMSE, SSIM, and variance attained the values of 24.71, 236.67, 54.96 dB, 0.22, 63.16, and 0.11, respectively. This innovative method offers a promising direction for enhancing image fusion quality and is highlighted by its unique integration of optimisation and guidance strategies.

Quantum dots boost large-view NIR-II imaging with high fidelity for fluorescence-guided tumor surgery

Owing to the high spatiotemporal resolution, the second near-infrared (NIR-II) imaging window can provide high imaging contrast with diminished tissue autofluorescence and suppressed photon scattering to pinpoint the locations for tumor surgery. Due to the unique optical properties and excellent fluorescence performance, quantum dots (QDs) are regarded as ideal nanoprobes for fluorescence-guided surgery (FGS). Moreover, QDs can be excited by a variety of light sources owing to the continuous and wide absorption ranges. Herein, light-emitting diode (LED) was used as the excitation source of QDs-based nanoprobes to realize FGS of tumor with high resolution. Since the LED light could irradiate a large region with consistent light intensity, signal distortion at the edge of imaging field was avoided. The signal intensity of the view edges under LED excitation can be improved by about 5 times compared to laser excitation. Therefore, more micro-vessels and smaller tumors (Vtumor < 5 mm2) could be detected, thus providing more precise guidance for tumor resection surgery.

Unveiling the relationship between social anxiety, loneliness, motivations, and problematic smartphone use: A network approach

BackgroundPrevious studies suggested social anxiety as an essential risk factor for problematic smartphone use, but the complex interactions and the most influential components affecting this relationship remain unclear. This study capitalizes on network analysis to identify the central factors and possible mediating paths among social anxiety, loneliness, five types of motivation, and problematic smartphone use. Material and methodsEmploying 549 emerging adults, we obtained a stable network of the above variables. The central components and the stability of this network were also identified. ResultsWithin this network, the edge linking withdrawal behavior and use of application (APP) exhibits the most robust edge intensity. The central components include social comfort, use of APP, withdrawal behavior, and companionship while the bridge central nodes include social anxiety and escapism motivation. The direct link between social anxiety and PSU revealed only fragile edges with both withdrawal behavior and use of APP. Considering the possible mediating pathways, three pathways were observed in our network. Loneliness and escapism mediated the relationship between social anxiety and social comfort. Moreover, another mediating way was from social anxiety, loneliness, social interaction motivation, and escapism motivation to social comfort. DiscussionBased on the above identification of related components and pathways, future researchers could intervene against problematic smartphone usage in this socially anxious population.

Research on dense object detection methods in congested environments of urban streets and roads based on DCYOLO

The urban street is a congested environment that contains a large number of occluded and size-differentiated objects. Aiming at the problems of the loss of the target to be detected and low detection accuracy resulting from this situation, a newly improved algorithm, based on YOLOv4, DCYOLO is proposed. Firstly, a Difference sensitive network (DSN) is introduced to extract the edge features of objects from the original image. Then, assign the edge features back to increase the edge intensity of the object in the original image and ultimately improve the detection performance. Secondly, the feature fusion module (CFFB) based on context information is introduced to realize the cross-scale fusion of shallow fine-grained features and deep-level features, to strengthen the cross-scale semantic information fusion of feature maps and eventually improve the performance of object detection. At last, in the network prediction part, the SIOU loss function replaces the original CIOU loss function to improve the convergence speed and accuracy of object detection. The experiments based on MS COCO 2017 and self-made datasets show that, compared with the YOLOv4, the detection accuracy of DCYOLO models is greatly improved with an increase of 9.1 percentage points in AP and 10.4 percentage points in APs. Compared with YOLOv5x and Faster R-CNN, DCYOLO shows higher accuracy and better detection performance. The experiment result proves that the DCYOLO algorithm can adapt to the dense object detection requirements in the congested environment of urban streets.

Latent profile and network analysis of risk perception among a sample of Chinese university students during the COVID-19 pandemic: a cross-sectional and longitudinal study.

The risk perception of contracting COVID-19 is an important topic for assessing and predicting COVID-19 infection and health education during the pandemic. However, studies that use latent profiles and network analysis together to measure the risk perception of COVID-19 are rare. Therefore, this study combined latent profile analysis and network analysis to measure risk perception toward COVID-19 among Chinese university students through a cross-sectional and longitudinal study. A sample of 1,837 Chinese university students (735 males, 40%) completed the cross-sectional study with an eight-item risk perception questionnaire in January 2020, while 334 Chinese university students (111 males, 33.2%) completed the longitudinal study at three time points. A two-class model including a low risk perception class (n = 1,005, 54.7%) and a high risk perception class (n = 832, 45.3%) was selected for the cross-sectional study. Nodes rp6 ("Average people have chances of contracting COVID-19'') and rp7 ("Average people worry about catching COVID-19") had the strongest edge intensity (r = 0.491), while node rp5 ("The COVID-19 outbreak affects the whole country") had the highest strength centrality in the cross-sectional study. The risk perception of contracting COVID-19 decreased continuously at the three time points. Moreover, the network structures and global strengths had no significant differences in the longitudinal study. The risk perception of contracting COVID-19 decreased continually during the COVID-19 pandemic, which indicated the importance of cultural influence and effective government management in China. In addition, university students displayed strong trust and confidence in the government's ability to fight COVID-19. The results indicate that the government should take strong measures to prevent and intervene in various risks and reinforce the public's trust through positive media communications.

Hybrid Simulated Annealing‐Evaporation Rate‐Based Water Cycle Algorithm Application for Medical Image Enhancement

Visualizing medical images is difficult due to artifacts, poor local contrast, low soft tissue contrast, excessive noise levels, and a wide dynamic range. This has created a serious problem for physicians, resulting in unapproachable and inaccurate disease diagnoses. To circumvent this problem, this research proposes a medical image enhancement (MIE) approach based on the hybrid simulated annealing‐evaporation rate‐based water cycle algorithm (SA‐ERWCA). The ERWCA enhances distorted medical images by finding the best optimal solution for the transformation parameters according to the fitness function. The fitness function consists of three objective measurements, namely, entropy, number of edges, and sum of edge intensities. Due to its high potential for finding a globally optimal solution, SA is applied solely to determine the optimized initial population of the ERWCA, thereby enhancing its convergence characteristics. To simply put, the main objective of SA is to avoid premature convergence of the ERWCA. Thus, blending SA and ERWCA produces better‐quality medical images. The performance of the proposed algorithm was compared to histogram equalization (HE), low contrast stretching (LCS), contrast‐limited adaptive histogram equalization (CLAHE), particle swarm optimization (PSO), accelerated PSO (APSO), and the water cycle algorithm (WCA). Along with the objective function fitness, seven full reference (FR) image quality assessment (IQA) metrics were implemented to evaluate image quality and compare performance. The findings of the present study showed that the proposed strategy outperforms all of the compared methods in terms of objective function fitness and perceptual visual IQA metrics such as the Harr wavelet‐based perceptual similarity index (HaarPSI), visual signal‐to‐noise ratio (VSNR), and information content weighted structural similarity index measure (IW‐SSIM). The suggested technique also exhibited better stability and faster convergence time to the optimum solution. Furthermore, the proposed approach outperformed others by avoiding premature convergence to the best solution and providing excellent optimization results with acceptable computational efficiency.

Multi-scale Based Approach for Crater Detection on Lunar Surface using Clustering algorithm

This paper introduced automated detection and analysis of lunar craters are crucial for advancing lunar research and facilitating mission planning. This study introduces a comprehensive multi-scale approach for crater detection on the lunar surface using LROC (Lunar Reconnaissance Orbiter Camera) Dataset, enabling the identification of craters across varying scales and enhancing overall detection accuracy and efficiency. The process involves preprocessing lunar imagery to enhance features and delineate potential crater regions, utilizing edge detection algorithms to extract crater boundaries. Clustering techniques are then applied to group similar edge points and isolate potential crater candidates. A thresholding mechanism based on statistical edge intensity analysis refines the detection process. To address multiple detections of the same crater at slightly different diameters, radius criteria are implemented, yielding high diameter accuracy of 84% for mare Imbrium (test site 1) and 80% for mare Nubium (test site 2). Subsequently, a depth-diameter analysis validates crater-like characteristics by combining depth measurements with crater diameter estimation, showcasing the authenticity of potential craters. we compare our diameter and depth with different reference papers and we get high accuracy of R2 =0.99 then other methods. The results highlight the potential of this multi-scale approach for automated lunar crater detection, providing deeper insights into the Moon's surface morphology and history for future lunar exploration missions.

(Invited) Nanoscale Devices for Accelerating AI Algorithms

The energy consumption of training and deploying state-of-the-art artificial intelligence (AI) models has experienced exponential growth, driven by increasing model parameters and the necessary training data. Recent models demand over 1 billion PetaFLOPs of total computation for training, potentially taking weeks to complete. With current GPU performance at approximately 1-3 TOPS/W, GPU training energy consumption alone can exceed 100,000 kWh, equivalent to the monthly energy expenditure of 100 US households. Over the next decade, these models are expected to scale up further, driving total computing energy to constitute a significant portion of global consumption. In this talk, we describe both synaptic and neuronal devices that can accelerate AI algorithms with potentially multiple orders of magnitude improvement in power efficiency.First, we describe an oscillatory retinal neuron (ORN) that directly converts incident DC light into voltage spikes. Coupled arrays of this device result in an imager that carries out in-sensor processing while capturing an image. Uniquely, the conversion from input light to output voltage spikes occurs without external power. When coupled in arrays, the neighboring neurons interact with each other to influence the spiking frequency spectrum. This allows the arrays to carry out frequency multiplexed computation on an input image. It is shown that this approach can achieve>20,000 TOPS/W, multiple orders of magnitude greater than the current approaches. Theory and simulation is used to elucidate how coupled ORNs carry out computation on an input image. When coupled, each output frequency band encodes a unique computation on the input image. By tuning the coupling impedances and the frequency bands, user defined computations can be carried out on the input. We experimentally show that this can be carried out with a 3x3 array that demonstrates simultaneous edge detection, intensity filtering, image segmentation and other functions. This hardware is then used to demonstrate improvement in MNIST handwritten digit classification accuracy over a traditional imager connected to a fully connected network.Next, we describe spiking synaptic devices that can be directly fabricated in the back-end of line of CMOS devices. These devices consist of an InP transistor channel with an engineered gate stack. Using a uniform gate insulator, we can demonstrate behaviors of biological synapses, such as potentiation, depression, spike number dependent plasticity, and spike timing dependent plasticity. By introducing a heterostructured gate insulator, it is shown that short-term to long-term memory transitions can be designed into the device. Finally, by using a transparent gate, an in-sensor synaptic phototransistor is demonstrated and the performance of these devices at a system level is demonstrated.

Selective image segmentation driven by region, edge and saliency functions

Present active contour methods often struggle with the segmentation of regions displaying variations in texture, color, or intensity a phenomenon referred to as inhomogeneities. These limitation impairs their ability to precisely distinguish and outline diverse components within an image. Further some of these methods employ intricate mathematical formulations for energy minimization. Such complexity introduces computational sluggishness, making these methods unsuitable for tasks requiring real-time processing or rapid segmentation. Moreover, these methods are susceptible to being trapped in energy configurations corresponding to local minimum points. Consequently, the segmentation process fails to converge to the desired outcome. Additionally, the efficacy of these methods diminishes when confronted with regions exhibiting weak or subtle boundaries. To address these limitations comprehensively, our proposed approach introduces a fresh paradigm for image segmentation through the synchronization of region-based, edge-based, and saliency-based segmentation techniques. Initially, we adapt an intensity edge term based on the zero crossing feature detector (ZCD), which is used to highlight significant edges of an image. Secondly, a saliency function is formulated to detect salient regions from an image. We have also included a globally tuned region based SPF (signed pressure force) term to move contour away and capture homogeneous regions. ZCD, saliency and global SPF are jointly incorporated with some scaled value for the level set evolution to develop an effective image segmentation model. In addition, proposed method is capable to perform selective object segmentation, which enables us to choose any single or multiple objects inside an image. Saliency function and ZCD detector are considered feature enhancement tools, which are used to get important features of an image, so this method has a solid capacity to segment nature images (homogeneous or inhomogeneous) precisely. Finally, the adaption of the Gaussian kernel removes the need of any penalization term for level set reinitialization. Experimental results will exhibit the efficiency of the proposed method.