Fully Convolutional Network Model Research Articles

Comparison of Fully Convolutional Networks and U-Net for Optic Disc and Optic Cup Segmentation

Glaucoma, the leading cause of irreversible blindness, must be diagnosed early and thus treated in time. However, it has no noticeable symptoms in its early stages and may not be detected easily. This paper aims to compare two well-known convolutional neural network (CNN) structures, namely Fully Convolutional Networks (FCNs) and U-Net for the segmentation of the optic disc (OD) and optic cup (OC) from retinal fundus images which play an important role in glaucoma diagnosis. The performance of both models is assessed using qualitative parameters such as the Dice coefficient, Jaccard index, and cup-to-disc ratio (CDR) error. In our experiment, the U-Net model yields more accurate segmentation results with 0.9601 average pixel accuracy and 0.9255 dice score for OD segmentation, outperforming the FCNs model with 0.9560 average pixel accuracy and 0.9132 dice score for OD segmentation. However, FCNs have a shorter inference time of 0. 0043 seconds against U-net’s 0. 0062 seconds making FCNs more suitable for real-time applications. The restrictions related to this study include biases from using only one dataset acquired from particular imaging devices, dependency on mask-based cropping techniques, and comparison being restricted to two fundamental architectures. This work presents the contribution of the deep learning models in improving glaucoma screening and therefore helping in avoiding blindness.

Forecasting future earthquakes with deep neural networks: application to California

SUMMARY We use the spatial map of the logarithm of past estimated released earthquake energies as input of fully convolutional networks (FCN) to forecast future earthquakes. This model is applied to California and compared with an elaborated version of the epidemic type aftershock sequence (ETAS) model. Our long-term earthquake forecast simulations show that the FCN model is close to the ETAS model in forecasting earthquakes with $M \ge 3.0,\,\,4.0,\,\,{\rm{and\,\,}}5.0$ according to the Molchan diagram. Moreover, training and implementing the FCN model is 2000–4000 times faster than calibrating the ETAS model and generating its probabilistic forecasts. The FCN model is straightforward in terms of its neural network structure and feature engineering. It does not require extensive knowledge of statistical seismology or the analysis of earthquake catalogue completeness. Using the earthquake catalogue with $M \ge 0$ as FCN input can enhance the model's performance in some time–magnitude forecasting windows.

Efficiency in constraint: A comparative analysis review of FCN and DeepLab models on small-scale datasets

Semantic segmentation, crucial in computer vision, differentiates objects within images and finds applications in autonomous vehicles, medical imaging, and assistive technology. It typically employs neural networks for pixel-wise image classification. Key advancements in this field are attributed to Fully Convolutional Networks (FCN) and DeepLab models, known for their high performance with extensive datasets. However, the challenge arises when these models are applied to smaller datasets. Our research presents a concise overview of seminal work in machine learning and semantic segmentation, followed by an exploration of FCN and DeepLab architectures. The study primarily focuses on evaluating these models efficacy on a smaller dataset. Results, summarized in tables and figures, indicate that FCN-16 outperforms others in limited-data scenarios, while DeepLab shows reduced effectiveness. This finding is significant for applications with constrained data resources, providing a direction for future research in semantic segmentation under such conditions.

Can prediction of acute myocardial infarction using a deep learning-enabled electrocardiogram be used in real-world practice?

Abstract Background Diagnosing acute myocardial infarction (MI) in the emergency department poses challenges. The utilization of a 12-lead electrocardiogram (ECG) is pivotal for detecting ST-segment elevation myocardial infarction (STEMI), yet enhancing ECG interpretation through a systematic algorithm could significantly improve diagnostic accuracy. Nonetheless, delayed diagnosis or misdiagnosis of STEMI is not uncommon in daily practice. Purpose This study aims to develop a fully convolutional network (FCN) based on ECG data as a diagnostic support tool and to evaluate its effectiveness in STEMI detection. Methods A retrospective cohort study involved 3,298 patients with angiographically confirmed STEMI and 37,393 control ECGs, comprising normal readings, false positives, and ECGs mimicking STEMI. The FCN underwent internal training, external validation, and testing across various data splits (80:10:10, 70:15:15, and 60:20:20). Performance evaluation on validation data included assessing the area under the receiver operating characteristic curve (AUC). Subsequently, the trained model predicted unseen data, and sensitivity, specificity, false positives, and false negatives were calculated to evaluate model performance. Results The FCN exhibited exceptional performance in STEMI detection during both internal training and external validation across different data splits. The receiver operating characteristic curves yielded AUC values of 0.995 (80:10:10), 0.994 (70:15:15), and 0.995 (60:20:20). Consistently high values were observed in precision-recall curves, with AUC values of 0.974 (80:10:10), 0.962 (70:15:15), and 0.973 (60:20:20). When utilizing this FCN model in the clinical setting to reanalyze 20 cases of previously misdiagnosed electrocardiograms, it correctly diagnosed 17 cases. Conclusion Integrating the FCN into clinical practice can notably improve the timely and accurate diagnosis of acute MI, reducing instances of delayed or erroneous STEMI diagnoses and facilitating prompt reperfusion therapy.

The Application Research of FCN Algorithm in Different Severe Convection Short-Time Nowcasting Technology in China, Gansu Province

This study explores the application of the fully convolutional network (FCN) algorithm to the field of meteorology, specifically for the short-term nowcasting of severe convective weather events such as hail, convective wind gust (CG), thunderstorms, and short-term heavy rain (STHR) in Gansu. The training data come from the European Center for Medium-Range Weather Forecasts (ECMWF) and real-time ground observations. The performance of the proposed FCN model, based on 2017 to 2021 training datasets, demonstrated a high prediction accuracy, with an overall error rate of 16.6%. Furthermore, the model exhibited an error rate of 18.6% across both severe and non-severe weather conditions when tested against the 2022 dataset. Operational deployment in 2023 yielded an average critical success index (CSI) of 24.3%, a probability of detection (POD) of 62.6%, and a false alarm ratio (FAR) of 71.2% for these convective events. It is noteworthy that the predicting performance for STHR was particularly effective with the highest POD and CSI, as well as the lowest FAR. CG and hail predictions had comparable CSI and FAR scores, although the POD for CG surpassed that for hail. The FCN model’s optimal performances in terms of hail prediction occurred at the 4th, 8th, and 10th forecast hours, while for CG, the 6th hour was most accurate, and for STHR, the 2nd and 4th hours were most effective. These findings underscore the FCN model’s ideal suitability for short-term forecasting of severe convective weather, presenting extensive prospects for the automation of meteorological operations in the future.

Fully Convolutional Network Model Applied Attention Mechanism on Kitti Lane Dataset for Lane Detection

This article investigated the attention mechanism implemented by the Fully Convolutional Network (FCN) Model on the Kitti Lane Dataset. Two attention mechanisms were applied in the deep learning model to improve traffic lane detection for autonomous vehicles. The Kitti lane dataset, which was generated in collaboration with Jannik Fritsch and Tobias Kuehl from Honda Research Europe GmbH, was selected for this study. The results demonstrate that the applied attention mechanism can effectively improve the network's feature representation on lane markings. Furthermore, this approach can improve the weighted information of lane line targets while decreasing irrelevant information. As a result, the proposed technique improved, obtaining more than 95% accuracy. Subsequently, the attention mechanism was implemented in the FCN model architecture to enhance the lane-detecting model. As a result, in the future, more comprehensive ideas, such as combining the FCN model with Transfer Learning, will play an essential part in investigating the improvement of lane detection areas.

Investigation of medical image segmentation techniques and analysis of key applications

This research examines the application of the UNet convolutional neural network model, specifically for semantic segmentation tasks in the field of medical imaging, juxtaposing its efficacy with Fully Convolutional Networks (FCNs). The primary focus of this comparative analysis rests on the performance of the UNet model on the dataset employed for this study. Surpassing our initial expectations, the UNet model demonstrated remarkable performance superiority over the FCN model on the curated dataset, thereby suggesting its potential applicability and utility for analogous tasks within the realm of medical imaging. In a surprising turn of events, our trials revealed that data augmentation techniques did not usher in a notable enhancement in segmentation accuracy. This observation was especially striking given the substantial size of the dataset employed for the experiments, encompassing as many as 1000 images. This outcome suggests that the merits of data augmentation may not always come to the fore when dealing with considerably large datasets. This intriguing discovery prompts further exploration and investigation to uncover the underlying reasons behind this observed phenomenon. Moreover, it brings to light an open-ended research query - the quest for alternative methodologies that could potentially amplify segmentation accuracy when operating on large scale datasets in the sphere of medical imaging. As the field continues to evolve and mature, it is these open questions that will continue to push the boundaries of what is possible in medical image analysis.

RTC_TongueNet: An improved tongue image segmentation model based on DeepLabV3.

Tongue segmentation as a basis for automated tongue recognition studies in Chinese medicine, which has defects such as network degradation and inability to obtain global features, which seriously affects the segmentation effect. This article proposes an improved model RTC_TongueNet based on DeepLabV3, which combines the improved residual structure and transformer and integrates the ECA (Efficient Channel Attention Module) attention mechanism of multiscale atrous convolution to improve the effect of tongue image segmentation. In this paper, we improve the backbone network based on DeepLabV3 by incorporating the transformer structure and an improved residual structure. The residual module is divided into two structures and uses different residual structures under different conditions to speed up the frequency of shallow information mapping to deep network, which can more effectively extract the underlying features of tongue image; introduces ECA attention mechanism after concat operation in ASPP (Atrous Spatial Pyramid Pooling) structure to strengthen information interaction and fusion, effectively extract local and global features, and enable the model to focus more on difficult-to-separate areas such as tongue edge, to obtain better segmentation effect. The RTC_TongueNet network model was compared with FCN (Fully Convolutional Networks), UNet, LRASPP (Lite Reduced ASPP), and DeepLabV3 models on two datasets. On the two datasets, the MIOU (Mean Intersection over Union) and MPA (Mean Pixel Accuracy) values of the classic model DeepLabV3 were higher than those of FCN, UNet, and LRASPP models, and the performance was better. Compared with the DeepLabV3 model, the RTC_TongueNet network model increased MIOU value by 0.9% and MPA value by 0.3% on the first dataset; MIOU increased by 1.0% and MPA increased by 1.1% on the second dataset. RTC_TongueNet model performed best on both datasets. In this study, based on DeepLabV3, we apply the improved residual structure and transformer as a backbone to fully extract image features locally and globally. The ECA attention module is combined to enhance channel attention, strengthen useful information, and weaken the interference of useless information. RTC_TongueNet model can effectively segment tongue images. This study has practical application value and reference value for tongue image segmentation.

Data Augmentation for Classification of Multi-Domain Tension Signals

There are different deep neural network (DNN) architectures and methods for performing augmentation on time series data, but not all the methods can be adapted for specific datasets. This article explores the development of deep learning models for time series, applies data augmentation methods to conveyor belt (CB) tension signal data and investigates the influence of these methods on the accuracy of CB state classification. CB systems are one of the essential elements of production processes, enabling smooth transportation of various industrial items, therefore its analysis is highly important. For the purpose of this work, multi-domain tension data signals from five different CB load weight conditions (0.5 kg, 1 kg, 2 kg, 3 kg, 5 kg) and one damaged belt condition were collected and analysed. Four DNN models based on fully convolutional network (FCN), convolutional neural network combined with long short-term memory (CNN-LSTM) model, residual network (ResNet), and InceptionTime architectures were developed and applied to classification of CB states. Different time series augmentations, such as random Laplace noise, drifted Gaussian noise, uniform noise, and magnitude warping, were applied to collected data during the study. Furthermore, new CB tension signals were generated using a TimeVAE model. The study has shown that DNN models based on FCN, ResNet, and InceptionTime architectures are able to classify CB states accurately. The research has also shown that various data augmentation methods can improve the accuracy of the above-mentioned models, for example, the combined addition of random Laplace and drifted Gaussian noise improved FCN model’s baseline (without augmentation) classification accuracy with 2.0 s-length signals by 4.5% to 92.6% ± 1.54%. FCN model demonstrated the best accuracy and classification performance despite its lowest amount of trainable parameters, thus demonstrating the importance of selecting and optimizing the right architecture when developing models for specific tasks.

Predicting the wall-shear stress and wall pressure through convolutional neural networks

The objective of this study is to assess the capability of convolution-based neural networks to predict the wall quantities in a turbulent open channel flow, starting from measurements within the flow. Gradually approaching the wall, the first tests are performed by training a fully-convolutional network (FCN) to predict the two-dimensional velocity-fluctuation fields at the inner-scaled wall-normal location ytarget+, using the sampled velocity fluctuations in wall-parallel planes located farther from the wall, at yinput+. The predictions from the FCN are compared against the predictions from a proposed R-Net architecture as a part of the network investigation study. Since the R-Net model is found to perform better than the FCN model, the former architecture is optimized to predict the two-dimensional streamwise and spanwise wall-shear-stress components and the wall pressure from the sampled velocity-fluctuation fields farther from the wall. The data for training and testing is obtained from direct numerical simulation (DNS) of open channel flow at friction Reynolds numbers Reτ=180 and 550. The turbulent velocity-fluctuation fields are sampled at various inner-scaled wall-normal locations, i.e.y+={15,30,50,100,150}, along with the wall-shear stress and the wall pressure. At Reτ=550, both FCN and R-Net can take advantage of the self-similarity in the logarithmic region of the flow and predict the velocity-fluctuation fields at y+=50 using the velocity-fluctuation fields at y+=100 as input with about 10% error in prediction of streamwise-fluctuations intensity. Further, the network model trained in this work is also able to predict the wall-shear-stress and wall-pressure fields using the velocity-fluctuation fields at y+=50 with around 10% error in the intensity of the corresponding fluctuations at both Reτ=180 and 550. These results are an encouraging starting point to develop neural-network-based approaches for modelling turbulence near the wall in numerical simulations, especially large-eddy simulations (LESs).

Abdominal fat quantification using convolutional networks

ObjectivesTo present software for automated adipose tissue quantification of abdominal magnetic resonance imaging (MRI) data using fully convolutional networks (FCN) and to evaluate its overall performance—accuracy, reliability, processing effort, and time—in comparison with an interactive reference method.Materials and methodsSingle-center data of patients with obesity were analyzed retrospectively with institutional review board approval. Ground truth for subcutaneous (SAT) and visceral adipose tissue (VAT) segmentation was provided by semiautomated region-of-interest (ROI) histogram thresholding of 331 full abdominal image series. Automated analyses were implemented using UNet-based FCN architectures and data augmentation techniques. Cross-validation was performed on hold-out data using standard similarity and error measures.ResultsThe FCN models reached Dice coefficients of up to 0.954 for SAT and 0.889 for VAT segmentation during cross-validation. Volumetric SAT (VAT) assessment resulted in a Pearson correlation coefficient of 0.999 (0.997), relative bias of 0.7% (0.8%), and standard deviation of 1.2% (3.1%). Intraclass correlation (coefficient of variation) within the same cohort was 0.999 (1.4%) for SAT and 0.996 (3.1%) for VAT.ConclusionThe presented methods for automated adipose-tissue quantification showed substantial improvements over common semiautomated approaches (no reader dependence, less effort) and thus provide a promising option for adipose tissue quantification.Clinical relevance statementDeep learning techniques will likely enable image-based body composition analyses on a routine basis. The presented fully convolutional network models are well suited for full abdominopelvic adipose tissue quantification in patients with obesity.Key Points• This work compared the performance of different deep-learning approaches for adipose tissue quantification in patients with obesity.• Supervised deep learning–based methods using fully convolutional networks were suited best.• Measures of accuracy were equal to or better than the operator-driven approach.

Classification of cow behavior patterns using inertial measurement units and a fully convolutional network model

In this study, we aimed to classify 7 cow behavior patterns automatically with an inertial measurement unit (IMU) using a fully convolutional network (FCN) algorithm. Behavioral data of 12 cows were collected by attaching an IMU in a waterproof box on the neck behind the head of each cow. Seven behavior patterns were considered: rub scratching (leg), ruminating-lying, lying, feeding, self-licking, rub scratching (neck), and social licking. To simplify the data and compare classification performance with or without magnetometer data, the 9-axis IMU data were reduced using the square root of the sum of squares to develop 2 datasets. Comparing the classification accuracy of the 3 models using a window size of 64 with 6-axis data and a window size of 128 with both 6-axis and 9-axis data, the best overall accuracy (83.75%) was achieved using the FCN model with a window size of 128 (12.8 s) using all IMU data. This model achieved classification accuracies of 83.2, 96.5, 92.8, 98.1, 82.9, 87.2, and 45.2% for ruminating-lying, lying, feeding, rub scratching (leg), self-licking, rub scratching (neck), and social licking, respectively. As a sequence of varied and intensive movement, the classification accuracy of behavior patterns related to skin disease was lower; better classification of these behavior patterns could be achieved with full IMU data and a larger window size. In the future, additional data will take into account different data types, such as audio and video data, to further enhance performance. In addition, an adaptive sliding window size will be used to improve model performance.

Segmentation of liver tumors with abdominal computed tomography using fully convolutional networks.

Dividing liver organs or lesions depicting on computed tomography (CT) images could be applied to help tumor staging and treatment. However, most existing image segmentation technologies use manual or semi-automatic analysis, making the analysis process costly and time-consuming. This research aims to develop and apply a deep learning network architecture to segment liver tumors automatically after fine tuning parameters. The medical imaging is obtained from the International Symposium on Biomedical Imaging (ISBI), which includes 3D abdominal CT scans of 131 patients diagnosed with liver tumors. From these CT scans, there are 7,190 2D CT images along with the labeled binary images. The labeled binary images are regarded as gold standard for evaluation of the segmented results by FCN (Fully Convolutional Network). The backbones of FCN are extracted from Xception, InceptionresNetv2, MobileNetv2, ResNet18, ResNet50 in this study. Meanwhile, the parameters including optimizers (SGDM and ADAM), size of epoch, and size of batch are investigated. CT images are randomly divided into training and testing sets using a ratio of 9:1. Several evaluation indices including Global Accuracy, Mean Accuracy, Mean IoU (Intersection over Union), Weighted IoU and Mean BF Score are applied to evaluate tumor segmentation results in the testing images. The Global Accuracy, Mean Accuracy, Mean IoU, Weighted IoU, and Mean BF Scores are 0.999, 0.969, 0.954, 0.998, 0.962 using ResNet50 in FCN with optimizer SGDM, batch size 12, and epoch 9. It is important to fine tuning the parameters in FCN model. Top 20 FNC models enable to achieve higher tumor segmentation accuracy with Mean IoU over 0.900. The occurred frequency of InceptionresNetv2, MobileNetv2, ResNet18, ResNet50, and Xception are 9, 6, 3, 5, and 2 times. Therefore, the InceptionresNetv2 has higher performance than others. This study develop and test an automated liver tumor segmentation model based on FCN. Study results demonstrate that many deep learning models including InceptionresNetv2, MobileNetv2, ResNet18, ResNet50, and Xception have high potential to segment liver tumors from CT images with accuracy exceeding 90%. However, it is still difficult to accurately segment tiny and small size tumors by FCN models.

Deep Learning Application for Classification of Ionospheric Height Profiles Measured by Radio Occultation Technique

Modern space missions provide a great number of height profiles of ionospheric electron density, measured by the remote sensing technique of radio occultation (RO). The deducing of the profiles from the RO measurements suffers from bias, resulting in negative values of the electron density. We developed a machine learning technique that allows automatic identification of ionospheric layers and avoids the bias problem. An algorithm of convolutional neural networks was applied for the classification of the height profiles. Six classes of the profiles were distinguished on the base of prominent ionospheric layers F2, Es, E, F1 and F3, as well as distorted profiles (Sc). For the models, we selected the ground truth of more than 712 height profiles measured by the COSMIC/Formosat-3 mission above Taiwan from 2011 to 2013. Two different models, a 1D convolutional neural network (CNN) and fully convolutional network (FCN), were applied for classification. It was found that both models demonstrate the best classification performance, with the average accuracy around 0.8 for prediction of the F2 layer-related class and the E layer-related class. The F1 layer is classified by the models with good performance (>0.7). The CNN model can effectively classify the Es layer with an accuracy of 0.75. The FCN model has good classification performance (0.72) for the Sc-related profiles. The lowest performance (<0.4) was found for the F3 layer-related class. It was shown that the more complex FCN model has better classification performance for both large-scale and small-scale variations in the height profiles of the ionospheric electron density.

Spectral index-driven FCN model training for water extraction from multispectral imagery

Terrestrial water is a fundamental component of the land surface. Accurate and robust water delineation of surface water is rather challenging due to the high intra-class variability and the spectral similarity with shadows and other dark surfaces. This study proposes a new method called the water index-driven deep fully convolutional neural network (WIDFCN) for high-accuracy water delineation with no need to collect samples manually. We formulate water delineation as a semantic labeling problem and solve it by training a deep fully convolutional network (FCN), whose capability of effectively extracting multilevel spatial and spectral features is exploited for discriminating water from complex surroundings. The main obstacle of using FCN, which requires a large volume of labeled training samples, is settled by utilizing the water recognition ability of the water spectral index (WI). Specifically, the training samples are automatically generated from WI by extracting a high-precision but incomplete water mask at first, which is then expanded to enhance the completeness. This strategy ensures the high quality of the automatically generated training samples and thus the water extraction performance of the trained FCN model. Twelve test sites from Sentinel-2 imagery with various water delineation challenges all over the world are used to assess the performance relative to that of supervised and unsupervised classification methods and water spectral index thresholding methods, in terms of Kappa coefficient, precision, recall, and F1-score. Overall, WIDFCN can achieve the highest precision and comparable recall, leading to the best water delineation accuracy with Kappa coefficient 0.9673 and F1-score 0.9696, as well as the lowest fluctuation in terms of various test sites. The results further demonstrate that WIDFCN can effectively deal with the scale and spectra variance of surface water, and has distinct robustness with respect to different kinds of shadows, including building, mountain, and cloud shadows. The findings in this study demonstrate a novel, robust, low-cost, and manual labor-free water delineation method that performs well in terms of both precision and completeness. Moreover, the core idea could provide a reference for extraction of geographic information by utilizing FCN models with no needs of manually labeling costs. Code and data will be available at https://github.com/LZhenShi/WIDFCN.

Complex-Valued Multi-Scale Fully Convolutional Network with Stacked-Dilated Convolution for PolSAR Image Classification

Polarimetric synthetic aperture radar (PolSAR) image classification is a pixel-wise issue, which has become increasingly prevalent in recent years. As a variant of the Convolutional Neural Network (CNN), the Fully Convolutional Network (FCN), which is designed for pixel-to-pixel tasks, has obtained enormous success in semantic segmentation. Therefore, effectively using the FCN model combined with polarimetric characteristics for PolSAR image classification is quite promising. This paper proposes a novel FCN model by adopting complex-valued domain stacked-dilated convolution (CV-SDFCN). Firstly, a stacked-dilated convolution layer with different dilation rates is constructed to capture multi-scale features of PolSAR image; meanwhile, the sharing weight is employed to reduce the calculation burden. Unfortunately, the labeled training samples of PolSAR image are usually limited. Then, the encoder–decoder structure of the original FCN is reconstructed with a U-net model. Finally, in view of the significance of the phase information for PolSAR images, the proposed model is trained in the complex-valued domain rather than the real-valued domain. The experiment results show that the classification performance of the proposed method is better than several state-of-the-art PolSAR image classification methods.

Detection of Patients with Congenital and Often Concealed Long-QT Syndrome by Novel Deep Learning Models.

Introduction: The long-QT syndrome (LQTS) is the most common ion channelopathy, typically presenting with a prolonged QT interval and clinical symptoms such as syncope or sudden cardiac death. Patients may present with a concealed phenotype making the diagnosis challenging. Correctly diagnosing at-risk patients is pivotal to starting early preventive treatment. Objective: Identification of congenital and often concealed LQTS by utilizing novel deep learning network architectures, which are specifically designed for multichannel time series and therefore particularly suitable for ECG data. Design and Results: A retrospective artificial intelligence (AI)-based analysis was performed using a 12-lead ECG of genetically confirmed LQTS (n = 124), including 41 patients with a concealed LQTS (33%), and validated against a control cohort (n = 161 of patients) without known LQTS or without QT-prolonging drug treatment but any other cardiovascular disease. The performance of a fully convolutional network (FCN) used in prior studies was compared with a different, novel convolutional neural network model (XceptionTime). We found that the XceptionTime model was able to achieve a higher balanced accuracy score (91.8%) than the associated FCN metric (83.6%), indicating improved prediction possibilities of novel AI architectures. The predictive accuracy prevailed independently of age and QTc parameters. Conclusions: In this study, the XceptionTime model outperformed the FCN model for LQTS patients with even better results than in prior studies. Even when a patient cohort with cardiovascular comorbidities is used. AI-based ECG analysis is a promising step for correct LQTS patient identification, especially if common diagnostic measures might be misleading.

Deep Implicit Statistical Shape Models for 3D Medical Image Delineation

3D delineation of anatomical structures is a cardinal goal in medical imaging analysis. Prior to deep learning, statistical shape models (SSMs) that imposed anatomical constraints and produced high quality surfaces were a core technology. Today’s fully-convolutional networks (FCNs), while dominant, do not offer these capabilities. We present deep implicit statistical shape models (DISSMs), a new approach that marries the representation power of deep networks with the benefits of SSMs. DISSMs use an implicit representation to produce compact and descriptive deep surface embeddings that permit statistical models of anatomical variance. To reliably fit anatomically plausible shapes to an image, we introduce a novel rigid and non-rigid pose estimation pipeline that is modelled as a Markov decision process (MDP). Intra-dataset experiments on the task of pathological liver segmentation demonstrate that DISSMs can perform more robustly than four leading FCN models, including nnU-Net + an adversarial prior: reducing the mean Hausdorff distance (HD) by 7.5-14.3 mm and improving the worst case Dice-Sørensen coefficient (DSC) by 1.2-2.3%. More critically, cross-dataset experiments on an external and highly challenging clinical dataset demonstrate that DISSMs improve the mean DSC and HD by 2.1-5.9% and 9.9-24.5 mm, respectively, and the worst-case DSC by 5.4-7.3%. Supplemental validation on a highly challenging and low-contrast larynx dataset further demonstrate DISSM’s improvements. These improvements are over and above any benefits from representing delineations with high-quality surfaces.

Cleanup Sketched Drawings: Deep Learning-Based Model.

Rough drawings provide artists with a simple and efficient way to express shapes and ideas. Artists frequently use sketches to highlight their envisioned curves, using several groups' raw strokes. These rough sketches need enhancement to remove some subtle impurities and completely simplify curves over the sketched images. This research paper proposes using a fully convolutional network (FCNN) model to simplify rough raster drawings using deep learning. As input, the FCNN takes a sketch image of any size and automatically generates a high-quality simplified sketch image as output. Our model intuitively addresses the shortcomings in the rough sketch image, such as noises and unwanted background, as well as the low resolution of the rough sketch image. The FCNN model is trained by three raster image datasets, which are publicly available online. This paper demonstrates the efficiency and effectiveness of using deep learning in cleaning and improving the roughly drawn image in an automatic way. For evaluating the results, the mean squared error (MSE) metric was used. From experimental results, it was observed that an enhanced FCNN model reported better accuracy, reducing the prediction error by 0.08 percent for simplifying the rough sketch compared to the existing methods.

Characterization of Semantic Segmentation Models on Mobile Platforms for Self-Navigation in Disaster-Struck Zones

The role of unmanned vehicles for searching and localizing the victims in disaster impacted areas such as earthquake-struck zones is getting more important. Self-navigation on an earthquake zone has a unique challenge of detecting irregularly shaped obstacles such as road cracks, debris on the streets, and water puddles. In this paper, we characterize a number of state-of-the-art Fully Convolutional Network (FCN) models on mobile embedded platforms for self-navigation at these sites containing extremely irregular obstacles. We evaluate the models in terms of accuracy, performance, and energy efficiency. We present a few optimizations for our designed vision system. Lastly, we discuss the trade-offs of these models for a couple of mobile platforms that can each perform self-navigation. To enable vehicles to safely navigate earthquake-struck zones, we compile a new annotated image database of various earthquake impacted regions that is different than traditional road damage databases. We train our database with a number of state-of-the-art semantic segmentation models in order to identify obstacles unique to earthquake-struck zones. Based on the statistics and tradeoffs, an optimal FCN model is selected and applied to the mobile vehicular platforms. To our best knowledge, this is the first study that identifies unique challenges and discusses the accuracy, performance, and energy impact of edge-based self-navigation mobile vehicles for earthquake-struck zones. Our proposed database and trained models are publicly available.