CT screening has been proven to be effective for diagnosing lung cancer at its early manifestation in the form of pulmonary nodules, thus decreasing mortality. However, exponential increase image data makes their accurate assessment a very challenging task given that number radiologists is limited and they have overworked. Recently, numerous methods, especially ones based on deep learning with convolutional neural network (CNN), developed automatically detect classify nodules medical images....

Various deep convolutional neural networks (CNNs) have been used to distinguish between benign and malignant pulmonary nodules using CT images. However, single learner usually presents unsatisfied performance due limited hypothesis space, or falling into local minima, wrong selection of space. To tackle these issues, we propose build ensemble learners through fusing multiple CNN for classification. image patches 743 are extracted from LIDC-IDRI database utilized. First, eight with different...

Early and automatic detection of pulmonary nodules from CT lung screening is the prerequisite for precise management cancer. However, a large number false positives appear in order to increase sensitivity, especially detecting micro-nodules (diameter < 3 mm), which increases radiologists' workload causes unnecessary anxiety patients. To decrease positive rate, we propose use CNN models discriminate between non-nodules image patches.A total 13,179 21,315 marked by radiologists are extracted...

Numerous automatic systems of pulmonary nodules detection have been proposed, but very few them consecrated to micro-nodules (diameter < 3 mm) even though they are regarded as the earliest manifestations lung cancer. Moreover, most available present high false positive rate resulting from their incapability discriminating between and non-nodules. Thus, this paper proposes a system differentiate non-nodules in computed tomography (CT) images by an ensemble learning multiple-view 3-D...

Full waveform inversion (FWI) is a powerful tool for estimating the underground velocity model. However, it computationally expensive and resulting models tend to be not accurate enough. Thus, improve efficiency accuracy of FWI, we propose super-resolution (SR) method based on deep learning enhance resolution seismic Since edge images model are also widely used in geophysics, multitask (MTL) network with hard parameter sharing applied perform SR its images. The proposed MTL dubbed M-RUDSR...

Ovarian cancer is one of the three most common types gynecological globally, with high-grade serous ovarian being and aggressive histological type. Guided treatment typically involves platinum-based combination chemotherapy, necessitating assessment whether patient platinum resistant. This study proposes a deep learning-based method to determine resistant using multimodal positron emission tomography/computed tomography images. In total, 289 patients were included in this study. An...

Structural curvatures are widely used seismic attributes that help interpreters to understand both structural and stratigraphic features. Traditional curvature extractions mainly calculated from dip estimations through lateral scanning of events, which is not only a very time-costing approach but also influenced by parameter settings, frequency, data quality. In this article, we propose deep learning-based volumetric extraction directly derives volumes the response. To realize above...

Recently, a multitask learning framework named M: multitask, R: global residual skip connection structure, U: encoder–decoder structure of U-Net, D: dense and SR: super-resolution (M-RUDSR) has successfully improved the accuracy full-waveform inversion (FWI) results by enhancing resolution seismic velocity model. However, M-RUDSR does not make full use data even though it contains high wavenumber information, which can help enhance Moreover, effects employing realized simply increasing...

The outcome of endoscopic tasks can be significantly affected by the presence specular reflections. Although numerous methods have been proposed for reflection detection and suppression from images, they are inefficient, usually require tedious empirical parameters selection, perform poorly when handling large regions. To this end, we propose a robust efficient deep learning framework termed EasySpec identifying suppressing reflections images. Our consists two stages: stage stage. former is...

The traditional method to improve the resolution in electromagnetic inversion is increasing number of iterations, which displays poor non-linear mapping and strong non-uniqueness. To meet this challenge, a new strategy proposed via reconstructing geoelectric model for traitional results through deep neural networks (DNN). DNN possesses advantage on establishing an uncertain between low-resolution images high-resolution target images. In order recover high-precision model, we propose...

Deep learning techniques can help minimize inter-physician analysis variability and the medical expert workloads, thereby enabling more accurate diagnoses. However, their implementation requires large-scale annotated dataset whose acquisition incurs heavy time human-expertise costs. Hence, to significantly annotation cost, this study presents a novel framework that enables deployment of deep methods in ultrasound (US) image segmentation requiring only very limited manually samples. We...

Active contour-based image segmentation can be a very challenging task due to many factors such as high intensity inhomogeneity, presence of noise, complex shape, weak boundaries objects, and dependence on the position initial contour. We propose level set-based active contour method segment shape objects from images corrupted by noise inhomogeneity. The energy function proposed results combining global information local with some regularization factors. First, term is based scheme...

Although echocardiography plays a crucial role in cardiovascular disease diagnosis, the presence of speckle noise images can hinder their diagnostic effectiveness. Recent studies have focused on using deep learning models to reduce echocardiography. However, these models' performances tend be limited real data as they are usually trained with synthesized due unavailability noise-free images. Moreover, although Noise2Noise algorithm which is widely adopted for natural image denoising used...

Background: Ovarian cancer is among the three most frequent gynecologic cancers globally. High-grade serous ovarian (HGSOC) common and aggressive histological type. Guided treatment for HGSOC typically involves platinum-based combination chemotherapy, necessitating an assessment of whether patient platinum-resistant. The purpose this study to propose a deep learning-based method determine platinum-resistant using multimodal positron emission tomography/computed tomography (PET/CT) images....

Active contour models (ACM) have been proven to be the most promising model in solving different problems encountered image segmentation. This paper proposes a new region-based active for level set formulation which energy function is formulated using both local and global intensity fitting terms. The generalized Gaussian distribution has used as kernel of binary information. evolution equation consists three terms: term, term regularization term. We introduced Laplace operator into...

The success of modern deep learning algorithms for image segmentation heavily relies on the availability high-quality labels training. However, obtaining accurate is time-consuming and tedious, requires expertise. If directly trained with dataset noisy annotations, networks can easily overfit to result in poor performance, which might lead serious misinterpretation. To this end, we propose a pixel estimation approach based neural network, helps correct annotations resulting better prediction...

Deep learning algorithms have been widely applied in the field of medical image analysis. However, it is very difficult and time-consuming to obtain a large number accurate samples their corresponding labels. neural networks can be train without samples, which may lead unsatisfactory performance. Currently, synthesizing an effective solution this problem. One commonly adopted methods for labeled ultrasound images simulation-based approach, such as Field II software. Although these...

Deep learning models can enable more accurate and efficient segmentation of cardiac structures in echocardiography (echo). However, their success depends on the availability large-scale annotated training data, whose achievement is highly challenging medical imaging. In this paper, we propose a novel data augmentation approach to tackle scarcity automatic segmentation. The proposed termed DF-Aug (Disentanglement Fusion Augmentation), consists two main steps: image disentanglement fusion....