Deep Convolutional Neural Network Method Research Articles

Image quality evaluation in deep-learning-based CT noise reduction using virtual imaging trial methods: Contrast-dependent spatial resolution.

Deep-learning-based image reconstruction and noise reduction methods (DLIR) have been increasingly deployed in clinical CT. Accurate image quality assessment of these methods is challenging as the performance measured using physical phantoms may not represent the true performance of DLIR in patients since DLIR is trained mostly on patient images. In this work, we aim to develop a patient-data-based virtual imaging trial framework and, as a first application, use it to measure the spatial resolution properties of a DLIR method. The patient-data-based virtual imaging trial framework consists of five steps: (1) insertion of lesions into projection domain data using the acquisition geometry of the patient exam to simulate different lesion characteristics; (2) insertion of noise into projection domain data using a realistic photon statistical model of the CT system to simulate different dose levels; (3) creation of DLIR-processed images from projection or image data; (4) creation of ensembles of DLIR-processed patient images from a large number of noise and lesion realizations; and (5) evaluation of image quality using ensemble DLIR images. This framework was applied to measure the spatial resolution of a ResNet based deep convolutional neural network (DCNN) trained on patient images. Lesions in a cylindrical shape and different contrast levels (-500, -100, -50, -20, -10 HU) were inserted to the lower right lobe of the liver in a patient case. Multiple dose levels were simulated (50%, 25%, 12.5%). Each lesion and dose condition had 600 noise realizations. Multiple reconstruction and denoising methods were used on all the noise realizations, including the original filtered-backprojection (FBP), iterative reconstruction (IR), and the DCNN method with three different strength setting (DCNN-weak, DCNN-medium, and DCNN-strong). Mean lesion signal was calculated by performing ensemble averaging of all the noise realizations for each lesion and dose condition and then subtracting the lesion-present images from the lesion absent images. Modulation transfer functions (MTFs) both in-plane and along the z-axis were calculated based on the mean lesion signals. The standard deviations of MTFs at each condition were estimated with bootstrapping: randomly sampling (with replacement) all the DLIR/FBP/IR images from the ensemble data (600 samples) at each condition. The impact of varying lesion contrast, dose levels, and denoising strengths were evaluated. Statistical analysis with paired t-test was used to compare the z-axis and in-plane spatial resolution of five algorithms for five different contrasts and three dose levels. The in-plane and z-axis spatial resolution degradation of DCNN becomes more severe as the contrast or radiation dose decreased, or DCNN denoising strength increased. In comparison with FBP, a 59.5% and 4.1% reduction of in-plane and z-axis MTF (in terms of spatial frequencies at 50% MTF), respectively, was observed at low contrast (-10 HU) for DCNN with the highest denoising strength at 25% routine dose level. When the dose level reduces from 50% to 12.5% of routine dose, the in-plane and z-axis MTFs reduces from 92.1% to 76.3%, and from 98.9% to 95.5%, respectively, at contrast of -100 HU, using FBP as the reference. For most conditions of contrasts and dose levels, significant differences were found among the five algorithms, with the following relationship in both in-plane and cross-plane spatial resolution: FBP>DCNN-Weak>IR>DCNN-Medium>DCNN-Strong. The spatial resolution difference among algorithms decreases at higher contrast or dose levels. A patient-data-based virtual imaging trial framework was developed and applied to measuring the spatial resolution properties of a DCNN noise reduction method at different contrast and dose levels using real patient data. As with other non-linear image reconstruction and post-processing techniques, the evaluated DCNN method degraded the in-plane and z-axis spatial resolution at lower contrast levels, lower radiation dose, and higher denoising strength.

DeepMCGCN: Multi-channel Deep Graph Neural Networks

Graph neural networks (GNNs) have shown powerful capabilities in modeling and representing graph structural data across various graph learning tasks as an emerging deep learning approach. However, most existing GNNs focus on single-relational graphs and fail to fully utilize the rich and diverse relational information present in real-world graph data. In addition, deeper GNNs tend to suffer from overfitting and oversmoothing issues, leading to degraded model performance. To deeply excavate the multi-relational features in graph data and strengthen the modeling and representation abilities of GNNs, this paper proposes a multi-channel deep graph convolutional neural network method called DeepMCGCN. It constructs multiple relational subgraphs and adopts multiple GCN channels to learn the characteristics of different relational subgraphs separately. Cross-channel connections are utilized to obtain interactions between different relational subgraphs, which can learn node embeddings richer and more discriminative than single-channel GNNs. Meanwhile, it alleviates overfitting issues of deep models by optimizing convolution functions and adding residual connections between and within channels. The DeepMCGCN method is evaluated on three real-world datasets, and the experimental results show that its node classification performance outperforms that of single-channel GCN and other benchmark models, which improves the modeling and representation capabilities of the model.

Deep Convolution Neural Network to solve Problems for Appel Leaf Disease Detection

Orchardists are now concerned about apple leaf infections since they could result in crop failure. It is challenging for growers to identify the type of illness on apple leaves due to the large variety of diseases that can affect apple leaves. In this study, we cover four different illness categories: healthy, numerous diseased, rusty, and scabby. a deep convolutional neural network method of processing. using a number of suggested methods, including data preprocessing and the pre-configured VGG-16 deep convolutional neural network (CNN) architecture. The Adam optimization model's beta 2 = 0.999 parameter value at Ephoch to 85/100 with an accuracy of 0.7582 and epsilon = 1e-07 parameter value at Ephoch to 32/100 with an accuracy of 0.7582 both produced the best accuracy outcomes.

Comparison of deep learning convolutional neural networks method with conventional volume-based morphometry measurement of hippocampal volume in Alzheimer's disease

Dementia is a spectrum of diseases characterised by a progressive and irreversible decline in cognitive function. Appropriate tools and references are essential for evaluating individuals' cognitive levels, especially hippocampal volume, as it is the commonly used biomarker in detecting Alzheimer's disease (AD). It is important to note that while there is no cure for dementia, early intervention and support can greatly improve the lives of those affected. Our ongoing AD research is being conducted to develop new treatments and improve our understanding of the disease by using voxel-based morphometry (VBM) to compare sensitivity and specificity with the HippoDeep toolbox. We validated AD's hippocampal volume compared to age-matched healthy controls (HC) based on the HippoDeep Model by comparing it with VBM as the reference standard. Significant differences between hippocampal volume in AD and HC have been detected using VBM and HippoDeep analysis.

Liver segmentation using Mnet for cirrhosis

Liver biopsies for diagnosing cirrhosis, the last stage of chronic liver disease was analysed using imaging and it could able to determine the severity level of the disease. Liver lesions can be segmented for evaluation of tumor burden, therapeutic strategy, prognosis, and follow-up on the efficacy of therapy. Automatic technologies for malignancy identification and segmentation are preferable because manual segmentation is a laborious process that is prone to error. We thus propose an M-Net approach to cirrhosis of the liver segmentation. The liver datasets are collected initially as raw data, which, due to the existence of noise, must be normalized before the investigation can begin. With the help of the Bilateral Filter and the Wavelet Transform, we can get rid of the noise and improve the de-noised image. The Gabor Filter is used for feature extraction. Hybrid Genetic Algorithm (HGA) is used to pick the best feature subsets for classification. Ensemble Deep Convolutional Neural Network (EDCNN) method is applied to the classification process. MATLAB, a simulation program, is used to conduct the entire inquiry. In terms of accuracy, our system has outperformed the most sophisticated automatic techniques.

3-D Tactile-Based Object Recognition for Robot Hands Using Force-Sensitive and Bend Sensor Arrays

Tactile sensing is a particularly important and challenging task for a modern robot to safely manipulate objects, interact with humans in a shared space, and provide various services. This paper presents a 3D tactile glove for robots with the combination of a piezoresistive-based force sensor array (412 sensors) covering the full hand and a resistive bend sensor array (5 sensors) on the back of five fingers. Deep learning-based CNN (Convolutional Neural Network) methods using the designed tactile glove are proposed for object recognition. In the experiment for recognizing 15 objects with a dexterous robot hand, an average classification accuracy of 93.67% has been achieved. Comparison experiments with three other typical classifiers (the Quadratic SVM, Weighted KNN, and Bagged Trees) and our CNN methods show an average recognition accuracy of 91.67% with the 3D tactile glove, revealing an accuracy improvement of 4.17% over only using the force sensor array. We further apply our 3D tactile glove and the multi-modal CNN to identify three other objects and demonstrate their generalization ability of tactile object recognition with an average success accuracy of 78.33%. The proposed 3D tactile glove can be further used in human-robot interactions, the design of prosthetics and humanoid robots, and for improving the intelligence level in brain-computer collaborative systems.

A convolutional neural network deep learning method for model class selection

AbstractThe response‐only model class selection capability of a novel deep convolutional neural network method is examined herein in a simple, yet effective, manner. Specifically, the responses from a unique degree of freedom along with their class information train and validate a one‐dimensional convolutional neural network. In doing so, the network selects the model class of new and unlabeled signals without the need of the system input information, or full system identification. An optional physics‐based algorithm enhancement is also examined using the Kalman filter to fuse the system response signals using the kinematics constraints of the acceleration and displacement data. Importantly, the method is shown to select the model class in slight signal variations attributed to the damping behavior or hysteresis behavior on both linear and nonlinear dynamic systems, as well as on a 3D building finite element model, providing a powerful tool for structural health monitoring applications.

Block-based compressive sensing in deep learning using AlexNet for vegetable classification

Vegetables can be distinguished according to differences in color, shape, and texture. The deep learning convolutional neural network (CNN) method is a technique that can be used to classify types of vegetables for various applications in agriculture. This study proposes a vegetable classification technique that uses the CNN AlexNet model and applies compressive sensing (CS) to reduce computing time and save storage space. In CS, discrete cosine transform (DCT) is applied for the sparsing process, Gaussian distribution for sampling, and orthogonal matching pursuit (OMP) for reconstruction. Simulation results on 600 images for four types of vegetables showed a maximum test accuracy of 98% for the AlexNet method, while the combined block-based CS using the AlexNet method produced a maximum accuracy of 96.66% with a compression ratio of 2×. Our results indicated that AlexNet CNN architecture and block-based CS in AlexNet can classify vegetable images better than previous methods.

Crack inspection in tunnel structures by fusing information from a 3D light detection and ranging and pan-tilt-zoom camera system

This study proposes a practical crack inspection method for tunnel structures by fusing measurements from two novel sensors: 3D light detection and ranging (LiDAR) and pan-tilt-zoom (PTZ) camera. The proposed method comprises three phases: prerequisite phase, real-time measurement phase, and inspection phase. First, the 3D LiDAR and the PTZ camera is combined to generate 3D point cloud data (PCD) with RGB data prior to extract the plane contained the crack. Second, the PTZ camera tracked the crack of interest at the pre-registered location. Specifically, several signal processing methods including the Hough transform and random sample consensus are addressed to effectively recognize the crack of interest. Then, the pan and tilt of the camera are controlled to track the crack of interest and record the crack at several magnifications of the zoom automatically because the optimal magnification value of the zoom is unknown in practical field applications. Third, crack propagation rate is estimated by calculating the length of the crack using information on the crack recorded in previous measurements. Cracks in a recorded image are extracted by addressing a multiscale multilevel mask deep convolutional neural network (MSML Mask DCNN) and pixel-based clustering method. The MSML Mask DCNN constructs distinct feature maps to effectively extract features of cracks and the pixel-based clustering method effectively eliminate cracks not of interest to accurately calculate the crack propagation rate for the crack of interest. Finally, the crack propagation rate was calculated through statistical analysis with several images at different zoom magnifications to secure high accuracy. Field experiments demonstrate that the proposed method is effective to inspect cracks. Systematic analysis on experiments also revealed that all methods deployed in the crack inspection enable ensures high accuracy and robustness in field applications. Consequently, the proposed crack inspection system and method provide a cost-effective and autonomous solution for crack inspection of tunnel structures.

Automatic detection and classification of nasopalatine duct cyst and periapical cyst on panoramic radiographs using deep convolutional neural networks

The aim of this study was to evaluate a deep convolutional neural network (DCNN) method for the detection and classification of nasopalatine duct cysts (NPDC) and periapical cysts (PAC) on panoramic radiographs. A total of 1,209 panoramic radiographs with 606 NPDC and 603 PAC were labeled with a bounding box and divided into training, validation, and test sets with an 8:1:1 ratio. The networks used were EfficientDet-D3, Faster R-CNN, YOLO v5, RetinaNet, and SSD. Mean average precision (mAP) was used to assess performance. Sixty images with no lesion in the anterior maxilla were added to the previous test set and were tested on 2 dentists with no training in radiology (GP) and on EfficientDet-D3. The performances were comparatively examined. The mAP for each DCNN was EfficientDet-D3 93.8%, Faster R-CNN 90.8%, YOLO v5 89.5%, RetinaNet 79.4%, and SSD 60.9%. The classification performance of EfficientDet-D3 was higher than that of the GPs' with accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of 94.4%, 94.4%, 97.2%, 94.6%, and 97.2%, respectively. The proposed method achieved high performance for the detection and classification of NPDC and PAC compared with the GPs and presented promising prospects for clinical application.

New Hybrid Graph Convolution Neural Network with Applications in Game Strategy

Deep convolutional neural networks (DCNNs) have enjoyed much success in many applications, such as computer vision, automated medical diagnosis, autonomous systems, etc. Another application of DCNNs is for game strategies, where the deep neural network architecture can be used to directly represent and learn strategies from expert players on different sides. Many game states can be expressed not only as a matrix data structure suitable for DCNN training but also as a graph data structure. Most of the available DCNN methods ignore the territory characteristics of both sides’ positions based on the game rules. Therefore, in this paper, we propose a hybrid approach to the graph neural network to extract the features of the model of game-playing strategies and fuse it into a DCNN. As a graph learning model, graph convolutional networks (GCNs) provide a scheme by which to extract the features in a graph structure, which can better extract the features in the relationship between the game-playing strategies. We validate the work and design a hybrid network to integrate GCNs and DCNNs in the game of Go and show that on the KGS Go dataset, the performance of the hybrid model outperforms the traditional DCNN model. The hybrid model demonstrates a good performance in extracting the game strategy of Go.

Differentiation of lung and breast cancer brain metastases: Comparison of texture analysis and deep convolutional neural networks.

Metastases are the most common neoplasm in the adult brain. In order to initiate the treatment, an extensive diagnostic workup is usually required. Radiomics is a discipline aimed at transforming visual data in radiological images into reliable diagnostic information. We aimed to examine the capability of deep learning methods to classify the origin of metastatic lesions in brain MRIs and compare the deep Convolutional Neural Network (CNN) methods with image texture based features. One hundred forty three patients with 157 metastatic brain tumors were included in the study. The statistical and texture based image features were extracted from metastatic tumors after manual segmentation process. Three powerful pre-trained CNN architectures and the texture-based features on both 2D and 3D tumor images were used to differentiate lung and breast metastases. Ten-fold cross-validation was used for evaluation. Accuracy, precision, recall, and area under curve (AUC) metrics were calculated to analyze the diagnostic performance. The texture-based image features on 3D volumes achieved better discrimination results than 2D image features. The overall performance of CNN architectures with 3D inputs was higher than the texture-based features. Xception architecture, with 3D volumes as input, yielded the highest accuracy (0.85) while the AUC value was 0.84. The AUC values of VGG19 and the InceptionV3 architectures were 0.82 and 0.81, respectively. CNNs achieved superior diagnostic performance in differentiating brain metastases from lung and breast malignancies than texture-based image features. Differentiation using 3D volumes as input exhibited a higher success rate than 2D sagittal images.

Revolutionizing network management with an AI-driven intrusion detection system

The creation of methods and models that can learn and make predictions or judgments based on such learning is artificial intelligence (AI). By combining an AI-driven intrusion detection system (IDS) with the BAT optimization method and a Deep Convolutional Neural Network (DCNN), we provide a revolutionary strategy to the revolutionize network management. Utilizing the advantages of deep learning and BAT optimization, the goal is to increase the efficiency of intrusion detection in the network management. Here, the classification effectiveness of the DCNN is increased by using the BAT optimization strategy. The suggested framework combines a DCNN model, which is excellent in pattern data collection, pre-processing by using normalization, and prediction tasks, with the BAT Optimized with Deep Convolutional Neural Network (BATO-DCNN) method, recognized for its capacity to identify optimum solutions in the challenging search spaces. The suggested method effectively tunes the DCNN using BAT optimization, leading to the better convergence and accuracy. The results of the research show that the recommended methodology performs better than traditional approaches in terms of accuracy, precision, F1-score, and recall measures. The results of this study support the current research in the area of network security and open the door for improved network management systems.

Self-trained deep convolutional neural network for noise reduction in CT.

Supervised deep convolutional neural network (CNN)-based methods have been actively used in clinical CT to reduce image noise. The networks of these methods are typically trained using paired high- and low-quality data from a massive number of patients and/or phantom images. This training process is tedious, and the network trained under a given condition may not be generalizable to patient images acquired and reconstructed under different conditions. We propose a self-trained deep CNN (ST_CNN) method for noise reduction in CT that does not rely on pre-existing training datasets. The ST_CNN training was accomplished using extensive data augmentation in the projection domain, and the inference was applied to the data itself. Specifically, multiple independent noise insertions were applied to the original patient projection data to generate multiple realizations of low-quality projection data. Then, rotation augmentation was adopted for both the original and low-quality projection data by applying the rotation angle directly on the projection data so that images were rotated at arbitrary angles without introducing additional bias. A large number of paired low- and high-quality images from the same patient were reconstructed and paired for training the ST_CNN model. No significant difference was found between the ST_CNN and conventional CNN models in terms of the peak signal-to-noise ratio and structural similarity index measure. The ST_CNN model outperformed the conventional CNN model in terms of noise texture and homogeneity in liver parenchyma as well as better subjective visualization of liver lesions. The ST_CNN may sacrifice the sharpness of vessels slightly compared to the conventional CNN model but without affecting the visibility of peripheral vessels or diagnosis of vascular pathology. The proposed ST_CNN method trained from the data itself may achieve similar image quality in comparison with conventional deep CNN denoising methods pre-trained on external datasets.

Convolutional Neural Networks for Automated Diagnosis of Diabetic Retinopathy in Fundus Images

Diabetic retinopathy (DR), a long-term complication of diabetes, is notoriously hard to detect in its early stages due to the fact that it only shows a subset of symptoms. Standard diagnostic procedures for DR now include OCT and digital fundus imaging. If digital fundus images alone could provide a reliable diagnosis, then eliminating the costly optical coherence tomography would be beneficial for all parties involved. Optometrists and their patients will find this useful. Using deep convolutional neural networks, we provide a novel approach to this problem. Our approach deviates from standard DCNN methods by exchanging typical max-pooling layers with fractional max-pooling ones. In order to collect more subtle information for categorisation, two such DCNNs, each with a different number of layers, are trained. To establish these limits, we use deep convolutional neural networks (DCNNs) and features extracted from picture metadata to train a support vector machine classifier. In our experiments, we used information from Kaggle's open DR detection database. We fed our model 34,124 training images, 1,000 validation examples, and 53,572 test images to train and test it. Each of the five classes in the proposed DR classifier corresponds to one of the steps in the DR process and is given a numeric value between 0 and 4. Experimental results show a higher identification rate (86.17%) than those found in the existing literature, indicating the suggested strategy may be effective. We have jointly developed an algorithm for machine learning and accompanying software, and we've named it Deep Retina. Images of the fundus acquired by the typical person using a portable ophthalmoscope may be instantly analyzed using our technology. This technology might be used for self-diagnosis, at-home care, and telemedicine.

Evaluating the ecological vulnerability of Chongqing using deep learning.

This study used deep learning to evaluate the ecological vulnerability of Chongqing, China, discuss the deep learning evaluations of ecological vulnerability, and generate vulnerability maps that support local ecological environment protection and governance decisions and provide reference for future studies. The information gain ratio was used to screen the influencing factors, selecting 16 factors that influence ecological vulnerability. Deep neural network (DNN) and convolutional neural network (CNN) methods were used for modeling, and two ecological vulnerability maps of the study area were generated. The results showed that the mean absolute error and root mean square error of the DNN and CNN models were relatively small, and the fitting accuracy was high. The area under the receiver operating characteristic curve of the CNN model was 0.926, which was better than that of the DNN model (0.888). Random forest was applied to calculate the importance of the influencing factors in the two models. Because the main factor was geological features, the relative ecological vulnerability was mainly affected by karst topography. Through the analysis of the ecological vulnerability map, the areas with higher vulnerability are the karst mountains of Dabashan, Wushan, and Qiyaoshan in the northeast and southeast, as well as the valley between mountains and cities in the center and west of the study area. According to the investigation of these areas, the primary ecological problems are low forest quality, structural irregularities caused by self-geological factors, severe desertification, and soil erosion. Human activity is also an important factor that causes ecological vulnerability in the study area. In conclusion, deep learning, particularly CNN models, can be used for ecological vulnerability assessments. The ecological vulnerability maps conformed to the basic cognition of field surveys and can provide references for other deep learning vulnerability studies. While the overall vulnerability of the study area is not high, ecological problems that lead to its vulnerability should be addressed by future ecological protection and management measures.

WRANet: wavelet integrated residual attention U-Net network for medical image segmentation

Medical image segmentation is crucial for the diagnosis and analysis of disease. Deep convolutional neural network methods have achieved great success in medical image segmentation. However, they are highly susceptible to noise interference during the propagation of the network, where weak noise can dramatically alter the network output. As the network deepens, it can face problems such as gradient explosion and vanishing. To improve the robustness and segmentation performance of the network, we propose a wavelet residual attention network (WRANet) for medical image segmentation. We replace the standard downsampling modules (e.g., maximum pooling and average pooling) in CNNs with discrete wavelet transform, decompose the features into low- and high-frequency components, and remove the high-frequency components to eliminate noise. At the same time, the problem of feature loss can be effectively addressed by introducing an attention mechanism. The combined experimental results show that our method can effectively perform aneurysm segmentation, achieving a Dice score of 78.99%, an IoU score of 68.96%, a precision of 85.21%, and a sensitivity score of 80.98%. In polyp segmentation, a Dice score of 88.89%, an IoU score of 81.74%, a precision rate of 91.32%, and a sensitivity score of 91.07% were achieved. Furthermore, our comparison with state-of-the-art techniques demonstrates the competitiveness of the WRANet network.

GPS signal acquisition based on deep convolutional neural network and post-correlation methods

GPS signal acquisition based on deep convolutional neural network and post-correlation methods

Accurate delineation of individual tree crowns in tropical forests from aerial RGB imagery using Mask R‐CNN

AbstractTropical forests are a major component of the global carbon cycle and home to two‐thirds of terrestrial species. Upper‐canopy trees store the majority of forest carbon and can be vulnerable to drought events and storms. Monitoring their growth and mortality is essential to understanding forest resilience to climate change, but in the context of forest carbon storage, large trees are underrepresented in traditional field surveys, so estimates are poorly constrained. Aerial photographs provide spectral and textural information to discriminate between tree crowns in diverse, complex tropical canopies, potentially opening the door to landscape monitoring of large trees. Here we describe a new deep convolutional neural network method, Detectree2, which builds on the Mask R‐CNN computer vision framework to recognize the irregular edges of individual tree crowns from airborne RGB imagery. We trained and evaluated this model with 3797 manually delineated tree crowns at three sites in Malaysian Borneo and one site in French Guiana. As an example application, we combined the delineations with repeat lidar surveys (taken between 3 and 6 years apart) of the four sites to estimate the growth and mortality of upper‐canopy trees. Detectree2 delineated 65 000 upper‐canopy trees across 14 km2 of aerial images. The skill of the automatic method in delineating unseen test trees was good (F1 score = 0.64) and for the tallest category of trees was excellent (F1 score = 0.74). As predicted from previous field studies, we found that growth rate declined with tree height and tall trees had higher mortality rates than intermediate‐size trees. Our approach demonstrates that deep learning methods can automatically segment trees in widely accessible RGB imagery. This tool (provided as an open‐source Python package) has many potential applications in forest ecology and conservation, from estimating carbon stocks to monitoring forest phenology and restoration.Python package available to install at https://github.com/PatBall1/Detectree2.

Contrast- and noise-dependent spatial resolution measurement for deep convolutional neural network-based noise reduction in CT using patient data.

Deep convolutional neural network (DCNN)-based noise reduction methods have been increasingly deployed in clinical CT. Accurate assessment of their spatial resolution properties is required. Spatial resolution is typically measured on physical phantoms, which may not represent the true performance of DCNN in patients as it is typically trained and tested with patient images and the generalizability of DNN to physical phantoms is questionable. In this work, we proposed a patient-data-based framework to measure the spatial resolution of DCNN methods, which involves lesion- and noise-insertion in projection domain, lesion ensemble averaging, and modulation transfer function measurement using an oversampled edge spread function from the cylindrical lesion signal. The impact of varying lesion contrast, dose levels, and CNN denoising strengths were investigated for a ResNet-based DCNN model trained using patient images. The spatial resolution degradation of DCNN reconstructions becomes more severe as the contrast or radiation dose decreased, or DCNN denoising strength increased. The measured 50%/10% MTF spatial frequencies of DCNN with highest denoising strength were (-500 HU:0.36/0.72 mm-1; -100 HU:0.32/0.65 mm-1; -50 HU:0.27/0.53 mm-1; -20 HU:0.18/0.36 mm-1; -10 HU:0.15/0.30 mm-1), while the 50%/10% MTF values of FBP were almost kept constant of 0.38/0.76 mm-1.