A decision map contains complete and clear information about the image to be fused, which is crucial to various image fusion issues, especially multi-focus image fusion. However, in order to get a satisfactory image f...A decision map contains complete and clear information about the image to be fused, which is crucial to various image fusion issues, especially multi-focus image fusion. However, in order to get a satisfactory image fusion effect, getting a decision map is very necessary and usually difficult to finish. In this letter, we address this problem with convolutional neural network(CNN), aiming to get a state-of-the-art decision map. The main idea is that the max-pooling of CNN is replaced by a convolution layer, the residuals are propagated backwards by gradient descent, and the training parameters of the individual layers of the CNN are updated layer by layer. Based on this, we propose a new all CNN(ACNN)-based multi-focus image fusion method in spatial domain. We demonstrate that the decision map obtained from the ACNN is reliable and can lead to high-quality fusion results. Experimental results clearly validate that the proposed algorithm can obtain state-of-the-art fusion performance in terms of both qualitative and quantitative evaluations.展开更多
In neuropathological diseases such as Alzheimer's Disease(AD),neuroimaging and Magnetic Resonance Imaging(MRI)play crucial roles in the realm of Artificial Intelligence of Medical Things(AIoMT)by leveraging edge i...In neuropathological diseases such as Alzheimer's Disease(AD),neuroimaging and Magnetic Resonance Imaging(MRI)play crucial roles in the realm of Artificial Intelligence of Medical Things(AIoMT)by leveraging edge intelligence resources.However,accurately classifying MRI scans based on neurodegenerative diseases faces challenges due to significant variability across classes and limited intra-class differences.To address this challenge,we propose a novel approach aimed at improving the early detection of AD through MRI imaging.This method integrates a Convolutional Neural Network(CNN)with a Cascade Attention Model(CAM-CNN).The CAM-CNN model outperforms traditional CNNs in AD classification accuracy and processing complexity.In this architecture,the attention mechanism is effectively implemented by utilizing two constraint cost functions and a cross-network with diverse pre-trained parameters for a two-stream architecture.Additionally,two new cost functions,Satisfied Rank Loss(SRL)and Cross-Network Similarity Loss(CNSL),are introduced to enhance collaboration and overall network performance.Finally,a unique entropy addition method is employed in the attention module for network integration,converting intermediate outcomes into the final prediction.These components are designed to work collaboratively and can be sequentially trained for optimal performance,thereby enhancing the effectiveness of AD stage classification and robustness to interference from MR images.Validation using the Kaggle dataset demonstrates the model's accuracy of 99.07%in multiclass classification,ensuring precise classification and early detection of all AD subtypes.Further validation across three feature categories with varying numbers confirms the robustness of the proposed approach,with deviations from the standard criteria of less than 1%.Applied in Alzheimer's patient care,this capability holds promise for enhancing value-based therapy and clinical decision-making.It aids in differentiating Alzheimer's patients from healthy individuals,thereby improving patient care and enabling more targeted therapies.展开更多
The convolutional neural network(CNN)is one of the main algorithms that is applied to deep transfer learning for classifying two essential types of liver lesions;Hemangioma and hepatocellular carcinoma(HCC).Ultrasound...The convolutional neural network(CNN)is one of the main algorithms that is applied to deep transfer learning for classifying two essential types of liver lesions;Hemangioma and hepatocellular carcinoma(HCC).Ultrasound images,which are commonly available and have low cost and low risk compared to computerized tomography(CT)scan images,will be used as input for the model.A total of 350 ultrasound images belonging to 59 patients are used.The number of images with HCC is 202 and 148,respectively.These images were collected from ultrasound cases.info(28 Hemangiomas patients and 11 HCC patients),the department of radiology,the University of Washington(7 HCC patients),the Atlas of ultrasound Germany(3 HCC patients),and Radiopedia and others(10 HCC patients).The ultrasound images are divided into 225,52,and 73 for training,validation,and testing.A data augmentation technique is used to enhance the validation performance.We proposed an approach based on ensembles of the best-selected deep transfer models from the on-the-shelf models:VGG16,VGG19,DenseNet,Inception,InceptionResNet,ResNet,and EfficientNet.After tuning both the feature extraction and the classification layers,the best models are selected.Validation accuracy is used for model tuning and selection.The accuracy,sensitivity,specificity and AUROC are used to evaluate the performance.The experiments are concluded in five stages.The first stage aims to evaluate the base model performance by training the on-the-shelf models.The best accu-racy obtained in the first stage is 83.5%.In the second stage,we augmented the data and retrained the on-the-shelf models with the augmented data.The best accuracy we obtained in the second stage was 86.3%.In the third stage,we tuned the feature extraction layers of the on-the-shelf models.The best accuracy obtained in the third stage is 89%.In the fourth stage,we fine-tuned the classification layer and obtained an accuracy of 93%as the best accuracy.In the fifth stage,we applied the ensemble approach using the best three-performing models and obtained an accuracy,specificity,sensitivity,and AUROC of 94%,93.7%,95.1%,and 0.944,respectively.展开更多
Layered metal dichalcogenides (LMDs) neuromorphic memristor devices offer a promising alternative toconventional von Neumann architectures, addressing speedand energy efficiency constraints. However, challenges remain...Layered metal dichalcogenides (LMDs) neuromorphic memristor devices offer a promising alternative toconventional von Neumann architectures, addressing speedand energy efficiency constraints. However, challenges remainin controlling resistive switching and operating voltage incrystalline LMD memristors due to environmental stabilization issues, which hinder neural network hardware development. Herein, we introduce an optimization method formemristor operation by controlling oxidation through ozonetreatment, creating a SnO_(x)/SnS_(2) resistive layer. These optimized memristors demonstrate low switching voltages (~1 V),rapid switching speeds (~20 ns), high switching ratios (10^(2)),and the ability to emulate synaptic weight plasticity. Crosssectional transmission electron microscopy and energy-dispersive X-ray spectroscopy identified defects and Ti conductive filaments in the resistive switching layer, contributingto uniform switching and minimized operating variation. Thedevice achieved 90% accuracy in MNIST handwritten recognition, and hardware-based image convolution was successfully implemented, showcasing the potential of SnO_(x)/SnS_(2)memristors for neuromorphic applications.展开更多
The plenoptic imaging technique provides a promising approach to the non-invasive three-dimensional measurement, especially for the high-temperature combustion diagnosis. We establish a light-field convolution imaging...The plenoptic imaging technique provides a promising approach to the non-invasive three-dimensional measurement, especially for the high-temperature combustion diagnosis. We establish a light-field convolution imaging model for diffusion flame in this work, considering the radiation transfer process inside the diffusion flame and the light transfer process inside the focused plenoptic camera together. The radiation transfer process is described by the radiation transfer equation and solved by the generalized source multi-flux method. Wave optics theory is adopted to describe the light transfer process, combining Fresnel diffraction and the phase conversion of the lens. The flame light-field image is obtained by the light-field convolution imaging model and adopted as the measurement signal to reconstruct three-dimensional temperature field. The inverse problem of temperature reconstruction is solved by the least square QR decomposition method. The simulative temperature reconstruction work is conducted, including the inverse analysis, the uncertainty analysis, and the measurement noise influence. All the results show that the proposed measurement method is available to reconstruct three-dimensional temperature with satisfactory accuracy and acceptable uncertainty. Both symmetric and asymmetric distributed temperature fields are investigated, and the reconstructed results prove the validity and universality of the measurement method.展开更多
As an emerging information device that adapts to development of the big data era,memristor has attracted much attention due to its advantage in processing massive data.However,the nucleation and growth of conductive f...As an emerging information device that adapts to development of the big data era,memristor has attracted much attention due to its advantage in processing massive data.However,the nucleation and growth of conductive filaments often exhibit randomness and instability,which undoubtedly leads to a wide and discrete range of switching parameters,damaging the electrical performance of device.In this work,a strategy of inserting carbon quantum dots(CQDs)into graphene oxide(GO)resistance layer is utilized to improve the stability of the switching parameters and the reliability of the device is improved.Compared with GO-based devices,GO/CQDs/GO-based devices exhibit a more stable resistance switching curve,low power,lower and more concentrated threshold voltage parameters with lower variation coefficient,faster switching speed,and more stable retention and endurance.The cause-inducing performance improvement may be attributed to the local electric field generated by CQDs in resistance switching that effectively guides the formation and rupture of conductive filaments,which optimizes the effective migration distance of Ag^(+),thereby improving the uniformity of resistance switching.Additionally,a convolutional neural network model is constructed to identify the CIFAR-10 data set,showing the high recognition accuracy of online and offline learning.The cross-kernel structure is used to further implement convolutional image processing through multiplication and accumulation operations.This work provides a solution to improve the performance of memristors,which can contribute to developing digital information processing.展开更多
Carbon quantum dots (CQDs) have been used in memristors due to their attractive optical and electronic properties, which are considered candidates for brain-inspired computing devices. In this work, the performance of...Carbon quantum dots (CQDs) have been used in memristors due to their attractive optical and electronic properties, which are considered candidates for brain-inspired computing devices. In this work, the performance of CQDs-based memristors is improved by utilizing nitrogen-doping. In contrast, nitrogen-doped CQDs (N-CQDs)-based optoelectronic memristors can be driven with smaller programming voltages (−0.6 to 0.7 V) and exhibit lower powers (78 nW/0.29 µW). The physical mechanism can be attributed to the reversible transition between C–N and C=N with lower binding energy induced by the electric field and the generation of photogenerated carriers by ultraviolet light irradiation, which adjusts the conductivity of the initial N-CQDs to implement resistance switching. Importantly, the convolutional image processing based on various cross kernels is efficiently demonstrated by stable multi-level storage properties. An N-CQDs-based optoelectronic reservoir computing implements impressively high accuracy in both no noise and various noise modes when recognizing the Modified National Institute of Standards and Technology (MNIST) dataset. It illustrates that N-CQDs-based memristors provide a novel strategy for developing artificial vision system with integrated in-memory sensor and computing.展开更多
基金supported by the National Natural Science Foundation of China(No.61174193)
文摘A decision map contains complete and clear information about the image to be fused, which is crucial to various image fusion issues, especially multi-focus image fusion. However, in order to get a satisfactory image fusion effect, getting a decision map is very necessary and usually difficult to finish. In this letter, we address this problem with convolutional neural network(CNN), aiming to get a state-of-the-art decision map. The main idea is that the max-pooling of CNN is replaced by a convolution layer, the residuals are propagated backwards by gradient descent, and the training parameters of the individual layers of the CNN are updated layer by layer. Based on this, we propose a new all CNN(ACNN)-based multi-focus image fusion method in spatial domain. We demonstrate that the decision map obtained from the ACNN is reliable and can lead to high-quality fusion results. Experimental results clearly validate that the proposed algorithm can obtain state-of-the-art fusion performance in terms of both qualitative and quantitative evaluations.
基金funded by the National Elites Foundation(No.711.5095).
文摘In neuropathological diseases such as Alzheimer's Disease(AD),neuroimaging and Magnetic Resonance Imaging(MRI)play crucial roles in the realm of Artificial Intelligence of Medical Things(AIoMT)by leveraging edge intelligence resources.However,accurately classifying MRI scans based on neurodegenerative diseases faces challenges due to significant variability across classes and limited intra-class differences.To address this challenge,we propose a novel approach aimed at improving the early detection of AD through MRI imaging.This method integrates a Convolutional Neural Network(CNN)with a Cascade Attention Model(CAM-CNN).The CAM-CNN model outperforms traditional CNNs in AD classification accuracy and processing complexity.In this architecture,the attention mechanism is effectively implemented by utilizing two constraint cost functions and a cross-network with diverse pre-trained parameters for a two-stream architecture.Additionally,two new cost functions,Satisfied Rank Loss(SRL)and Cross-Network Similarity Loss(CNSL),are introduced to enhance collaboration and overall network performance.Finally,a unique entropy addition method is employed in the attention module for network integration,converting intermediate outcomes into the final prediction.These components are designed to work collaboratively and can be sequentially trained for optimal performance,thereby enhancing the effectiveness of AD stage classification and robustness to interference from MR images.Validation using the Kaggle dataset demonstrates the model's accuracy of 99.07%in multiclass classification,ensuring precise classification and early detection of all AD subtypes.Further validation across three feature categories with varying numbers confirms the robustness of the proposed approach,with deviations from the standard criteria of less than 1%.Applied in Alzheimer's patient care,this capability holds promise for enhancing value-based therapy and clinical decision-making.It aids in differentiating Alzheimer's patients from healthy individuals,thereby improving patient care and enabling more targeted therapies.
基金funded by the Deanship of Scientific Research at Jouf University under Grant No.(DSR-2022-RG-0104).
文摘The convolutional neural network(CNN)is one of the main algorithms that is applied to deep transfer learning for classifying two essential types of liver lesions;Hemangioma and hepatocellular carcinoma(HCC).Ultrasound images,which are commonly available and have low cost and low risk compared to computerized tomography(CT)scan images,will be used as input for the model.A total of 350 ultrasound images belonging to 59 patients are used.The number of images with HCC is 202 and 148,respectively.These images were collected from ultrasound cases.info(28 Hemangiomas patients and 11 HCC patients),the department of radiology,the University of Washington(7 HCC patients),the Atlas of ultrasound Germany(3 HCC patients),and Radiopedia and others(10 HCC patients).The ultrasound images are divided into 225,52,and 73 for training,validation,and testing.A data augmentation technique is used to enhance the validation performance.We proposed an approach based on ensembles of the best-selected deep transfer models from the on-the-shelf models:VGG16,VGG19,DenseNet,Inception,InceptionResNet,ResNet,and EfficientNet.After tuning both the feature extraction and the classification layers,the best models are selected.Validation accuracy is used for model tuning and selection.The accuracy,sensitivity,specificity and AUROC are used to evaluate the performance.The experiments are concluded in five stages.The first stage aims to evaluate the base model performance by training the on-the-shelf models.The best accu-racy obtained in the first stage is 83.5%.In the second stage,we augmented the data and retrained the on-the-shelf models with the augmented data.The best accuracy we obtained in the second stage was 86.3%.In the third stage,we tuned the feature extraction layers of the on-the-shelf models.The best accuracy obtained in the third stage is 89%.In the fourth stage,we fine-tuned the classification layer and obtained an accuracy of 93%as the best accuracy.In the fifth stage,we applied the ensemble approach using the best three-performing models and obtained an accuracy,specificity,sensitivity,and AUROC of 94%,93.7%,95.1%,and 0.944,respectively.
基金supported by the National Natural Science Foundation of China (22175060 and 12304082)Shenzhen Science and Technology Program (JCYJ20220530160407016)+1 种基金the Natural Science Foundation of Hunan Province (2023JJ20001)the support from the U.S. National Science Foundation (2004251)。
文摘Layered metal dichalcogenides (LMDs) neuromorphic memristor devices offer a promising alternative toconventional von Neumann architectures, addressing speedand energy efficiency constraints. However, challenges remainin controlling resistive switching and operating voltage incrystalline LMD memristors due to environmental stabilization issues, which hinder neural network hardware development. Herein, we introduce an optimization method formemristor operation by controlling oxidation through ozonetreatment, creating a SnO_(x)/SnS_(2) resistive layer. These optimized memristors demonstrate low switching voltages (~1 V),rapid switching speeds (~20 ns), high switching ratios (10^(2)),and the ability to emulate synaptic weight plasticity. Crosssectional transmission electron microscopy and energy-dispersive X-ray spectroscopy identified defects and Ti conductive filaments in the resistive switching layer, contributingto uniform switching and minimized operating variation. Thedevice achieved 90% accuracy in MNIST handwritten recognition, and hardware-based image convolution was successfully implemented, showcasing the potential of SnO_(x)/SnS_(2)memristors for neuromorphic applications.
基金supported by the National Natural Science Foundation of China(Grant No.51976044)the National Science and Technology Major Project(Grant No.2017-V-0016-0069)the Foundation for Heilongjiang Touyan Innovation Team Program。
文摘The plenoptic imaging technique provides a promising approach to the non-invasive three-dimensional measurement, especially for the high-temperature combustion diagnosis. We establish a light-field convolution imaging model for diffusion flame in this work, considering the radiation transfer process inside the diffusion flame and the light transfer process inside the focused plenoptic camera together. The radiation transfer process is described by the radiation transfer equation and solved by the generalized source multi-flux method. Wave optics theory is adopted to describe the light transfer process, combining Fresnel diffraction and the phase conversion of the lens. The flame light-field image is obtained by the light-field convolution imaging model and adopted as the measurement signal to reconstruct three-dimensional temperature field. The inverse problem of temperature reconstruction is solved by the least square QR decomposition method. The simulative temperature reconstruction work is conducted, including the inverse analysis, the uncertainty analysis, and the measurement noise influence. All the results show that the proposed measurement method is available to reconstruct three-dimensional temperature with satisfactory accuracy and acceptable uncertainty. Both symmetric and asymmetric distributed temperature fields are investigated, and the reconstructed results prove the validity and universality of the measurement method.
基金supported by the National Key Research and Development Program of China(No.2022YFE0139100)International cooperative research project of Jiangsu province(No.BZ2022008)+3 种基金the Natural Science Foundation Project of China(No.62175028)the Program 111_2.0 in China(No.BP0719013)Leading Technology of Jiangsu Basic Research Plan(No.BK20192003)Postgraduate Research&Practice Innovation Program of Jiangsu Province,and the Fundamental Research Funds for the Central Universities(No.KYCX23_0254).
文摘As an emerging information device that adapts to development of the big data era,memristor has attracted much attention due to its advantage in processing massive data.However,the nucleation and growth of conductive filaments often exhibit randomness and instability,which undoubtedly leads to a wide and discrete range of switching parameters,damaging the electrical performance of device.In this work,a strategy of inserting carbon quantum dots(CQDs)into graphene oxide(GO)resistance layer is utilized to improve the stability of the switching parameters and the reliability of the device is improved.Compared with GO-based devices,GO/CQDs/GO-based devices exhibit a more stable resistance switching curve,low power,lower and more concentrated threshold voltage parameters with lower variation coefficient,faster switching speed,and more stable retention and endurance.The cause-inducing performance improvement may be attributed to the local electric field generated by CQDs in resistance switching that effectively guides the formation and rupture of conductive filaments,which optimizes the effective migration distance of Ag^(+),thereby improving the uniformity of resistance switching.Additionally,a convolutional neural network model is constructed to identify the CIFAR-10 data set,showing the high recognition accuracy of online and offline learning.The cross-kernel structure is used to further implement convolutional image processing through multiplication and accumulation operations.This work provides a solution to improve the performance of memristors,which can contribute to developing digital information processing.
基金financially supported by the National Key Research and Development Program of China(No.2022YFE0139100)International cooperative research project of Jiangsu province(No.BZ2022008)+3 种基金the National Natural Science Foundation Project of China(No.62175028)Program 111_2.0 in China(No.BP0719013)Leading Technology of Jiangsu Basic Research Plan(No.BK20192003)Postgraduate Research & Practice Innovation Program of Jiangsu Province,the Fundamental Research Funds for the Central Universities(No.KYCX23_0254).
文摘Carbon quantum dots (CQDs) have been used in memristors due to their attractive optical and electronic properties, which are considered candidates for brain-inspired computing devices. In this work, the performance of CQDs-based memristors is improved by utilizing nitrogen-doping. In contrast, nitrogen-doped CQDs (N-CQDs)-based optoelectronic memristors can be driven with smaller programming voltages (−0.6 to 0.7 V) and exhibit lower powers (78 nW/0.29 µW). The physical mechanism can be attributed to the reversible transition between C–N and C=N with lower binding energy induced by the electric field and the generation of photogenerated carriers by ultraviolet light irradiation, which adjusts the conductivity of the initial N-CQDs to implement resistance switching. Importantly, the convolutional image processing based on various cross kernels is efficiently demonstrated by stable multi-level storage properties. An N-CQDs-based optoelectronic reservoir computing implements impressively high accuracy in both no noise and various noise modes when recognizing the Modified National Institute of Standards and Technology (MNIST) dataset. It illustrates that N-CQDs-based memristors provide a novel strategy for developing artificial vision system with integrated in-memory sensor and computing.
