Sepsis poses a serious threat to health of children in pediatric intensive care unit.The mortality from pediatric sepsis can be effectively reduced through in-time diagnosis and therapeutic intervention.The bacillicul...Sepsis poses a serious threat to health of children in pediatric intensive care unit.The mortality from pediatric sepsis can be effectively reduced through in-time diagnosis and therapeutic intervention.The bacilliculture detection method is too time-consuming to receive timely treatment.In this research,we propose a new framework:a deep encoding network with cross features(CF-DEN)that enables accurate early detection of sepsis.Cross features are automatically constructed via the gradient boosting decision tree and distilled into the deep encoding network(DEN)we designed.The DEN is aimed at learning sufficiently effective representation from clinical test data.Each layer of the DEN fltrates the features involved in computation at current layer via attention mechanism and outputs the current prediction which is additive layer by layer to obtain the embedding feature at last layer.The framework takes the advantage of tree-based method and neural network method to extract effective representation from small clinical dataset and obtain accurate prediction in order to prompt patient to get timely treatment.We evaluate the performance of the framework on the dataset collected from Shanghai Children's Medical Center.Compared with common machine learning methods,our method achieves the increase on F1-score by 16.06%on the test set.展开更多
Cross-diffusion is a ubiquitous phenomenon in complex networks, but it is often neglected in the study of reaction–diffusion networks. In fact, network connections are often random. In this paper, we investigate patt...Cross-diffusion is a ubiquitous phenomenon in complex networks, but it is often neglected in the study of reaction–diffusion networks. In fact, network connections are often random. In this paper, we investigate pattern dynamics of random networks with cross-diffusion by using the method of network analysis and obtain a condition under which the network loses stability and Turing bifurcation occurs. In addition, we also derive the amplitude equation for the network and prove the stability of the amplitude equation which is also an effective tool to investigate pattern dynamics of the random network with cross diffusion. In the meantime, the pattern formation consistently matches the stability of the system and the amplitude equation is verified by simulations. A novel approach to the investigation of specific real systems was presented in this paper. Finally, the example and simulation used in this paper validate our theoretical results.展开更多
The present progress of visual-based detection of the diseased area of a malady plays an essential part in the medicalfield.In that case,the image proces-sing is performed to improve the image data,wherein it inhibits ...The present progress of visual-based detection of the diseased area of a malady plays an essential part in the medicalfield.In that case,the image proces-sing is performed to improve the image data,wherein it inhibits unintended dis-tortion of image features or it enhances further processing in various applications andfields.This helps to show better results especially for diagnosing diseases.Of late the early prediction of cancer is necessary to prevent disease-causing pro-blems.This work is proposed to identify lung cancer using lung computed tomo-graphy(CT)scan images.It helps to identify cancer cells’affected areas.In the present work,the original input image from Lung Image Database Consortium(LIDC)typically suffers from noise problems.To overcome this,the Gaborfilter used for image processing is highly enhanced.In the next stage,the Spherical Iterative Refinement Clustering(SIRC)algorithm identifies cancer-suspected areas on the CT scan image.This approach can help radiologists and medical experts recognize cancer diseases and syndromes so that serious progress can be avoided in the early stages.These new methods help to remove unwanted por-tions of the CT image and better utilization the image.The subspace extraction of features approach is beneficial for evaluating lung cancer.This paper introduces a novel approach called Contiguous Cross Propagation Neural Network that tends to locate regions afflicted by lung cancer using CT scan pictures(CCPNN).By using the feature values from the fourth step of the procedure,the proposed CCPNN tends to categorize the lesion in the lung nodular site.The efficiency of the suggested CCPNN approach is evaluated using classification metrics such as recall(%),precision(%),F-measure(percent),and accuracy(%).Finally,the incorrect classification ratios are determined to compare the trained networks’effectiveness,through these parameters of CCPNN,it obtains the outstanding per-formance of 98.06%and it has provided the lowest false ratio of 1.8%.展开更多
How to organize crossing social network resources on a higher level of integration and address them to users＇ desktops is an important difficult problem. Especially, there is a lack of efficient approaches to softwar...How to organize crossing social network resources on a higher level of integration and address them to users＇ desktops is an important difficult problem. Especially, there is a lack of efficient approaches to software architecture to build reusable system over the crossing social network, From the viewpoint of temporal logic XYZ/E, this paper proposes a kind of Architecture Descrip- tion Language about the Crossing Social Network system （CSN_ADL）, which can be used to depict the main key processes over the cross-social network system, and formally defines some key concepts, such as relation component, corelation component, override corelation connector, interaction connector, corelation network-oriented architecture, as well as system correctness, system activity, and system safety. Furthermore, some properties of correctness, activity, and safety under the flame CSN_ADL is discussed and depicted formally, which provides a formally theo- retical instruction for architecture reuses.展开更多
We investigate the problem of how to minimize the energy consumption in multi-hop Wireless Sensor Network (WSN),under the constraint of end-to-end reliability Quality of Seervice (QoS) requirement.Based on the investi...We investigate the problem of how to minimize the energy consumption in multi-hop Wireless Sensor Network (WSN),under the constraint of end-to-end reliability Quality of Seervice (QoS) requirement.Based on the investigation,we jointly consider the routing,relay selection and power allocation algorithm,and present a novel distributed cross-layer strategy using opportunistic relaying scheme for cooperative communication.The results show that under the same QoS requirement,the proposed cross-layer strategy performs better than other cross-layer cooperative communication algorithms in energy efficiency.We also investigated the impact of several parameters on the energy efficiency of the cooperative communication in WSNs,thus can be used to provide guidelines to decide when and how to apply cooperation for a given setup.展开更多
Lossy link is one of the unique characteristics in random-deployed sensor networks. We envision that robustness and reliability of routing cannot be ensured purely in network layer. Our idea is to enhance the performa...Lossy link is one of the unique characteristics in random-deployed sensor networks. We envision that robustness and reliability of routing cannot be ensured purely in network layer. Our idea is to enhance the performance of routing protocol by cross-layer interaction. We modified mint protocol, a routing protocol in TinyOS and proposed an enhanced version of mint called PA-mint. A transmission power control interface is added to network layer in PA-mint. When routing performance of the current network is not satisfied, PA-mint monotonically increases the transmission power via the interface we added. PA-mint is able to connect orphan nodes and robust to node mobility or key nodes failure. In the case that automatic request retransmission is employed, the number of retransmissions can be reduced by PA-mint. Results from experiments show that PA-mint increases the reliability and robustness of routing protocol by cross-layer interaction.展开更多
This paper presents a wireless sensor network (WSN) access control algorithm designed to minimize WSN node energy consumption. Based on slotted ALOHA protocol, this algorithm incorporates the power control of physical...This paper presents a wireless sensor network (WSN) access control algorithm designed to minimize WSN node energy consumption. Based on slotted ALOHA protocol, this algorithm incorporates the power control of physical layer, the transmitting probability of medium access control (MAC) layer, and the automatic repeat request (ARQ) of link layer. In this algorithm, a cross-layer optimization is preformed to minimizing the energy consuming per bit. Through theory deducing, the transmitting probability and transmitting power level is determined, and the relationship between energy consuming per bit and throughput per node is provided. Analytical results show that the cross-layer algorithm results in a significant energy savings relative to layered design subject to the same throughput per node, and the energy saving is extraordinary in the low throughput region.展开更多
Several protocols and schemes have been proposed to reduce energy consumption in Wireless Sensor Net-works (WSNs). In this paper we employ farcoopt, a cross layer design approach with the concept of coop-eration among...Several protocols and schemes have been proposed to reduce energy consumption in Wireless Sensor Net-works (WSNs). In this paper we employ farcoopt, a cross layer design approach with the concept of coop-eration among the nodes with best farthest neighbor scheme to increase the Quality of Service (QoS), reduce energy consumption, increases performance and end-to-end throughput. We present cooperative transmission to connect previously disconnect parts of a network thus overcoming the separation problem of multi-hop network. We show that this approach improves connectivity over 50% compared to multi-hop approaches and reduces the number of nodes necessary to provide full coverage of an area up to 35%. Simulation results show that on increase of data rates i.e. packet the network life time increases in farcoopt as compared to tra-ditional multi hop approach. The result of this analysis is presented in this work.展开更多
We report a synthesis of microporous organic nanotube networks（MONNs） by a combination of hyper cross-linking and molecular templating of core-shell bottlebrush copolymers. The intrabrush and interbrush cross-linkin...We report a synthesis of microporous organic nanotube networks（MONNs） by a combination of hyper cross-linking and molecular templating of core-shell bottlebrush copolymers. The intrabrush and interbrush cross-linking of polystyrene（PS） shell layer in the core-shell bottlebrush copolymers led to the formation of micropores and large-sized nanopores（meso/macrospores） in MONNs, respectively, while selective removal of polylactide（PLA） core layer generated mesoporous tubular structure. The size of PLA-templated mesoporous cores and porous structure both at micro-and meso-scale could be controlled by simple tuning of the ratio of core/shell or the PLA core fraction in the bottlebrush precursors. Moreover, the resultant MONNs showed a highly selective adsorption capacity for the positively charged dyes on the basis of multi-porosity and carboxylate group-rich structure. In addition, MONNs also exhibited effective performance in size-selective adsorption of biomacromolecules. This work represents a new avenue for the preparation of MONNs and also provides a new application for molecular bottlebrushes in nanotechnology.展开更多
The Distributed Network Performance Measurement Sys-tem provides functions to derive performance indices of networks and services, which are significant for Network Management System. To make these two systems coopera...The Distributed Network Performance Measurement Sys-tem provides functions to derive performance indices of networks and services, which are significant for Network Management System. To make these two systems cooperate, we realize this cross-system invocation platform, using Web Service, a mechanism which allows two systems to exchange data over the internet through publishing interfaces [1]. There are several mature Web Service frameworks, Apache Axis2, Apache CXF etc. In this paper we choose Apache Axis2 to achieve the objective that the Network Management System can invocate the net-work performance measurement functions via the Web Services.展开更多
The main research objective in wireless sensor networks (WSN) domain is to develop algorithms and protocols to ensure minimal energy consumption with maximum network lifetime. In this paper, we propose a novel design ...The main research objective in wireless sensor networks (WSN) domain is to develop algorithms and protocols to ensure minimal energy consumption with maximum network lifetime. In this paper, we propose a novel design for energy harvesting sensor node and cross-layered MAC protocol using three adjacent layers (Physical, MAC and Network) to economize energy for WSN. The basic idea behind our protocol is to re-energize the neighboring nodes using the radio frequency (RF) energy transmitted by the active nodes. This can be achieved by designing new energy harvesting sensor node and redesigning the MAC protocol. The results show that the proposed cross layer CL_EHSN improves the life time of the WSN by 40%.展开更多
Communication over wireless links identifies significant challenges for routing protocols operating. This paper proposes a Cross-layer design based Multipath Routing Protocol (CMRP) for mobile ad hoc networks, by mean...Communication over wireless links identifies significant challenges for routing protocols operating. This paper proposes a Cross-layer design based Multipath Routing Protocol (CMRP) for mobile ad hoc networks, by means of the node energy signal from the physical layer. The purpose is to optimize routing decision and path quality. The nodes’ mobility behavior is predicted using a notion of “Signal Fading Degree, SFD”. Especially, in combination of the IEEE 802.11e standard at the MAC layer, we determine that the IEEE 802.11e makes a significant contribution to performance improvement of CMRP. Performance evaluation of AODV in legacy 802.11 and CMRP in IEEE 802.11e shows that, as a function of speed of node mobility, a tremendous reduction achieved, in metrics such as the average end-to-end delay, route overhead, route discovery frequency, normalized routing load - almost more than 80%, 40%, 40%, and 40%. In the case of varying number of sessions, the reduction for route discovery frequency and normalized routing load are up to 70% and 80%.展开更多
This study evaluates the performance and reliability of a vision transformer (ViT) compared to convolutional neural networks (CNNs) using the ResNet50 model in classifying lung cancer from CT images into four categori...This study evaluates the performance and reliability of a vision transformer (ViT) compared to convolutional neural networks (CNNs) using the ResNet50 model in classifying lung cancer from CT images into four categories: lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), large cell carcinoma (LULC), and normal. Although CNNs have made significant advancements in medical imaging, their limited capacity to capture long-range dependencies has led to the exploration of ViTs, which leverage self-attention mechanisms for a more comprehensive global understanding of images. The study utilized a dataset of 748 lung CT images to train both models with standardized input sizes, assessing their performance through conventional metrics—accuracy, precision, recall, F1 score, specificity, and AUC—as well as cross entropy, a novel metric for evaluating prediction uncertainty. Both models achieved similar accuracy rates (95%), with ViT demonstrating a slight edge over ResNet50 in precision and F1 scores for specific classes. However, ResNet50 exhibited higher recall for LULC, indicating fewer missed cases. Cross entropy analysis showed that the ViT model had lower average uncertainty, particularly in the LUAD, Normal, and LUSC classes, compared to ResNet50. This finding suggests that ViT predictions are generally more reliable, though ResNet50 performed better for LULC. The study underscores that accuracy alone is insufficient for model comparison, as cross entropy offers deeper insights into the reliability and confidence of model predictions. The results highlight the importance of incorporating cross entropy alongside traditional metrics for a more comprehensive evaluation of deep learning models in medical image classification, providing a nuanced understanding of their performance and reliability. While the ViT outperformed the CNN-based ResNet50 in lung cancer classification based on cross-entropy values, the performance differences were minor and may not hold clinical significance. Therefore, it may be premature to consider replacing CNNs with ViTs in this specific application.展开更多
文摘Sepsis poses a serious threat to health of children in pediatric intensive care unit.The mortality from pediatric sepsis can be effectively reduced through in-time diagnosis and therapeutic intervention.The bacilliculture detection method is too time-consuming to receive timely treatment.In this research,we propose a new framework:a deep encoding network with cross features(CF-DEN)that enables accurate early detection of sepsis.Cross features are automatically constructed via the gradient boosting decision tree and distilled into the deep encoding network(DEN)we designed.The DEN is aimed at learning sufficiently effective representation from clinical test data.Each layer of the DEN fltrates the features involved in computation at current layer via attention mechanism and outputs the current prediction which is additive layer by layer to obtain the embedding feature at last layer.The framework takes the advantage of tree-based method and neural network method to extract effective representation from small clinical dataset and obtain accurate prediction in order to prompt patient to get timely treatment.We evaluate the performance of the framework on the dataset collected from Shanghai Children's Medical Center.Compared with common machine learning methods,our method achieves the increase on F1-score by 16.06%on the test set.
基金Project supported by the National Natural Science Foundation of China(Grant Nos.11272277,11572278,and 11572084)the Innovation Scientists and Technicians Troop Construction Projects of Henan Province,China(Grant No.2017JR0013)
文摘Cross-diffusion is a ubiquitous phenomenon in complex networks, but it is often neglected in the study of reaction–diffusion networks. In fact, network connections are often random. In this paper, we investigate pattern dynamics of random networks with cross-diffusion by using the method of network analysis and obtain a condition under which the network loses stability and Turing bifurcation occurs. In addition, we also derive the amplitude equation for the network and prove the stability of the amplitude equation which is also an effective tool to investigate pattern dynamics of the random network with cross diffusion. In the meantime, the pattern formation consistently matches the stability of the system and the amplitude equation is verified by simulations. A novel approach to the investigation of specific real systems was presented in this paper. Finally, the example and simulation used in this paper validate our theoretical results.
文摘The present progress of visual-based detection of the diseased area of a malady plays an essential part in the medicalfield.In that case,the image proces-sing is performed to improve the image data,wherein it inhibits unintended dis-tortion of image features or it enhances further processing in various applications andfields.This helps to show better results especially for diagnosing diseases.Of late the early prediction of cancer is necessary to prevent disease-causing pro-blems.This work is proposed to identify lung cancer using lung computed tomo-graphy(CT)scan images.It helps to identify cancer cells’affected areas.In the present work,the original input image from Lung Image Database Consortium(LIDC)typically suffers from noise problems.To overcome this,the Gaborfilter used for image processing is highly enhanced.In the next stage,the Spherical Iterative Refinement Clustering(SIRC)algorithm identifies cancer-suspected areas on the CT scan image.This approach can help radiologists and medical experts recognize cancer diseases and syndromes so that serious progress can be avoided in the early stages.These new methods help to remove unwanted por-tions of the CT image and better utilization the image.The subspace extraction of features approach is beneficial for evaluating lung cancer.This paper introduces a novel approach called Contiguous Cross Propagation Neural Network that tends to locate regions afflicted by lung cancer using CT scan pictures(CCPNN).By using the feature values from the fourth step of the procedure,the proposed CCPNN tends to categorize the lesion in the lung nodular site.The efficiency of the suggested CCPNN approach is evaluated using classification metrics such as recall(%),precision(%),F-measure(percent),and accuracy(%).Finally,the incorrect classification ratios are determined to compare the trained networks’effectiveness,through these parameters of CCPNN,it obtains the outstanding per-formance of 98.06%and it has provided the lowest false ratio of 1.8%.
基金Supported by the Fujian Province Science Research Foundation Grant (2009J01272)the Research Fund (type A) (JA09038) from the Education Department of Fujian Provincethe Humanities and Social Science Research Projects of the Ministry of Education (11YJA860028)
文摘How to organize crossing social network resources on a higher level of integration and address them to users＇ desktops is an important difficult problem. Especially, there is a lack of efficient approaches to software architecture to build reusable system over the crossing social network, From the viewpoint of temporal logic XYZ/E, this paper proposes a kind of Architecture Descrip- tion Language about the Crossing Social Network system （CSN_ADL）, which can be used to depict the main key processes over the cross-social network system, and formally defines some key concepts, such as relation component, corelation component, override corelation connector, interaction connector, corelation network-oriented architecture, as well as system correctness, system activity, and system safety. Furthermore, some properties of correctness, activity, and safety under the flame CSN_ADL is discussed and depicted formally, which provides a formally theo- retical instruction for architecture reuses.
基金Supported by the 100 Top-Talents Program of Chinese Academic of Sciences (No. 99M2008M02)
文摘We investigate the problem of how to minimize the energy consumption in multi-hop Wireless Sensor Network (WSN),under the constraint of end-to-end reliability Quality of Seervice (QoS) requirement.Based on the investigation,we jointly consider the routing,relay selection and power allocation algorithm,and present a novel distributed cross-layer strategy using opportunistic relaying scheme for cooperative communication.The results show that under the same QoS requirement,the proposed cross-layer strategy performs better than other cross-layer cooperative communication algorithms in energy efficiency.We also investigated the impact of several parameters on the energy efficiency of the cooperative communication in WSNs,thus can be used to provide guidelines to decide when and how to apply cooperation for a given setup.
基金Supported by National Natural Science Foundation of P. R. China (60374072, 60434030)
文摘Lossy link is one of the unique characteristics in random-deployed sensor networks. We envision that robustness and reliability of routing cannot be ensured purely in network layer. Our idea is to enhance the performance of routing protocol by cross-layer interaction. We modified mint protocol, a routing protocol in TinyOS and proposed an enhanced version of mint called PA-mint. A transmission power control interface is added to network layer in PA-mint. When routing performance of the current network is not satisfied, PA-mint monotonically increases the transmission power via the interface we added. PA-mint is able to connect orphan nodes and robust to node mobility or key nodes failure. In the case that automatic request retransmission is employed, the number of retransmissions can be reduced by PA-mint. Results from experiments show that PA-mint increases the reliability and robustness of routing protocol by cross-layer interaction.
文摘This paper presents a wireless sensor network (WSN) access control algorithm designed to minimize WSN node energy consumption. Based on slotted ALOHA protocol, this algorithm incorporates the power control of physical layer, the transmitting probability of medium access control (MAC) layer, and the automatic repeat request (ARQ) of link layer. In this algorithm, a cross-layer optimization is preformed to minimizing the energy consuming per bit. Through theory deducing, the transmitting probability and transmitting power level is determined, and the relationship between energy consuming per bit and throughput per node is provided. Analytical results show that the cross-layer algorithm results in a significant energy savings relative to layered design subject to the same throughput per node, and the energy saving is extraordinary in the low throughput region.
文摘Several protocols and schemes have been proposed to reduce energy consumption in Wireless Sensor Net-works (WSNs). In this paper we employ farcoopt, a cross layer design approach with the concept of coop-eration among the nodes with best farthest neighbor scheme to increase the Quality of Service (QoS), reduce energy consumption, increases performance and end-to-end throughput. We present cooperative transmission to connect previously disconnect parts of a network thus overcoming the separation problem of multi-hop network. We show that this approach improves connectivity over 50% compared to multi-hop approaches and reduces the number of nodes necessary to provide full coverage of an area up to 35%. Simulation results show that on increase of data rates i.e. packet the network life time increases in farcoopt as compared to tra-ditional multi hop approach. The result of this analysis is presented in this work.
基金financially supported by the National Natural Science Foundation of China (Nos. 51273066 and 21574042)Shanghai Pujiang Program (No. 13PJ1402300)
文摘We report a synthesis of microporous organic nanotube networks（MONNs） by a combination of hyper cross-linking and molecular templating of core-shell bottlebrush copolymers. The intrabrush and interbrush cross-linking of polystyrene（PS） shell layer in the core-shell bottlebrush copolymers led to the formation of micropores and large-sized nanopores（meso/macrospores） in MONNs, respectively, while selective removal of polylactide（PLA） core layer generated mesoporous tubular structure. The size of PLA-templated mesoporous cores and porous structure both at micro-and meso-scale could be controlled by simple tuning of the ratio of core/shell or the PLA core fraction in the bottlebrush precursors. Moreover, the resultant MONNs showed a highly selective adsorption capacity for the positively charged dyes on the basis of multi-porosity and carboxylate group-rich structure. In addition, MONNs also exhibited effective performance in size-selective adsorption of biomacromolecules. This work represents a new avenue for the preparation of MONNs and also provides a new application for molecular bottlebrushes in nanotechnology.
文摘The Distributed Network Performance Measurement Sys-tem provides functions to derive performance indices of networks and services, which are significant for Network Management System. To make these two systems cooperate, we realize this cross-system invocation platform, using Web Service, a mechanism which allows two systems to exchange data over the internet through publishing interfaces [1]. There are several mature Web Service frameworks, Apache Axis2, Apache CXF etc. In this paper we choose Apache Axis2 to achieve the objective that the Network Management System can invocate the net-work performance measurement functions via the Web Services.
文摘The main research objective in wireless sensor networks (WSN) domain is to develop algorithms and protocols to ensure minimal energy consumption with maximum network lifetime. In this paper, we propose a novel design for energy harvesting sensor node and cross-layered MAC protocol using three adjacent layers (Physical, MAC and Network) to economize energy for WSN. The basic idea behind our protocol is to re-energize the neighboring nodes using the radio frequency (RF) energy transmitted by the active nodes. This can be achieved by designing new energy harvesting sensor node and redesigning the MAC protocol. The results show that the proposed cross layer CL_EHSN improves the life time of the WSN by 40%.
文摘Communication over wireless links identifies significant challenges for routing protocols operating. This paper proposes a Cross-layer design based Multipath Routing Protocol (CMRP) for mobile ad hoc networks, by means of the node energy signal from the physical layer. The purpose is to optimize routing decision and path quality. The nodes’ mobility behavior is predicted using a notion of “Signal Fading Degree, SFD”. Especially, in combination of the IEEE 802.11e standard at the MAC layer, we determine that the IEEE 802.11e makes a significant contribution to performance improvement of CMRP. Performance evaluation of AODV in legacy 802.11 and CMRP in IEEE 802.11e shows that, as a function of speed of node mobility, a tremendous reduction achieved, in metrics such as the average end-to-end delay, route overhead, route discovery frequency, normalized routing load - almost more than 80%, 40%, 40%, and 40%. In the case of varying number of sessions, the reduction for route discovery frequency and normalized routing load are up to 70% and 80%.
文摘This study evaluates the performance and reliability of a vision transformer (ViT) compared to convolutional neural networks (CNNs) using the ResNet50 model in classifying lung cancer from CT images into four categories: lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), large cell carcinoma (LULC), and normal. Although CNNs have made significant advancements in medical imaging, their limited capacity to capture long-range dependencies has led to the exploration of ViTs, which leverage self-attention mechanisms for a more comprehensive global understanding of images. The study utilized a dataset of 748 lung CT images to train both models with standardized input sizes, assessing their performance through conventional metrics—accuracy, precision, recall, F1 score, specificity, and AUC—as well as cross entropy, a novel metric for evaluating prediction uncertainty. Both models achieved similar accuracy rates (95%), with ViT demonstrating a slight edge over ResNet50 in precision and F1 scores for specific classes. However, ResNet50 exhibited higher recall for LULC, indicating fewer missed cases. Cross entropy analysis showed that the ViT model had lower average uncertainty, particularly in the LUAD, Normal, and LUSC classes, compared to ResNet50. This finding suggests that ViT predictions are generally more reliable, though ResNet50 performed better for LULC. The study underscores that accuracy alone is insufficient for model comparison, as cross entropy offers deeper insights into the reliability and confidence of model predictions. The results highlight the importance of incorporating cross entropy alongside traditional metrics for a more comprehensive evaluation of deep learning models in medical image classification, providing a nuanced understanding of their performance and reliability. While the ViT outperformed the CNN-based ResNet50 in lung cancer classification based on cross-entropy values, the performance differences were minor and may not hold clinical significance. Therefore, it may be premature to consider replacing CNNs with ViTs in this specific application.
