Purpose To evaluate the feasibility of leveraging serial low-dose CT (LDCT) scans to develop a radiomics-based reinforcement learning (RRL) model for improving early diagnosis lung cancer at baseline screening. Materials and Methods In this retrospective study, 1951 participants (female patients, 822; median age, 61 years [range, 55-74 years]) (male 1129; 62 were randomly selected from National Lung Screening Trial between August 2002 April 2004. An RRL using LDCT (S-RRL) was trained...

Automatic detection of myopia plays a significant role in clinical practice. Few studies have been done on the pathological myopia, and no attention has paid to distinguishment between it high myopia. Additionally, they are hard differentiate because similarity them. In this paper, we design network with two branches for different classification tasks, where first one is distinguish normal abnormal while other classify We manage improve accuracy by combining Binary Cross-Entropy loss Triplet...

OBJECTIVES: To evaluate the efficacy of treatments for aqueous misdirection syndrome and explore possible risk factors influencing prognosis. METHODS: Data including demographics, initial clinical characteristics ocular outcomes at follow-up were collected patients treated syndrome. Main outcome measures were: best-corrected visual acuity (BCVA); intraocular pressure (IOP); number antiglaucoma medications; recurrence; complications. RESULTS: available 50 (57 eyes). Final mean BCVA improved...

Accurate segmentation of the lung nodule in computed tomography images is a critical component computer-assisted cancer detection/diagnosis system. However, challenging task due to heterogeneity nodules. This study develop hybrid deep learning (H-DL) model for nodules with wide variety sizes, shapes, margins, and opacities.

Abstract Domestic animal embryonic stem (ES) cells would provide an invaluable research tool for genetic breeding and the production of transgenic animals. Unfortunately, authentic domestic animals ES have not been established despite progress made over more than two decades. Here, we show that ovine ES‐like can be efficiently derived propagated in a semi‐defined medium contains N2, B27, GSK3 inhibitor (CHIR99021), basic fibroblast growth factor (bFGF). These had characteristic...

Abstract Background Given the reliability of circRNAs in symbolizing cancer progression, this investigation was designed to expound involvement hsa_circ_0028007 regulating chemosensitivity nasopharyngeal carcinoma (NPC) cells. Methods Altogether, 241 pairs NPC tissues and para‐cancerous normal were collected identify NPC‐symbolic circRNAs, which have been screened by circRNA microarray advance. Expressions determined means real‐time polymerase chain reaction (PCR). Besides, human cell lines...

Introduction Recent efforts have been made to apply machine learning and deep approaches the automated classification of schizophrenia using structural magnetic resonance imaging (sMRI) at individual level. However, these are less accurate on early psychosis (EP) since there mild brain changes stage. As cognitive impairments is one main feature in psychosis, this study we a multi-task framework sMRI with inclusion assessment facilitate patients EP from healthy individuals. Method Unlike...

We study and address the cross-modal retrieval problem which lies at heart of visual-textual processing. Its major challenge in how to effectively learn a shared multi-modal feature space where discrepancies semantically related pairs, such as images texts, are minimized regardless their modalities. Most current methods focus on reasoning about cross-modality semantic relations within individual image-text pair common representation. However, they overlook more global, structural inter-pair...

BACKGROUND:Although resveratrol has been found to show anti-cancer effects and potential chemotherapeutic activities in several cancers, the role molecular mechanisms of nasopharyngeal carcinoma (NPC) remains poorly understood. This study aimed investigate effect NPC progression its mechanism. MATERIAL AND METHODS:Quantitative real-time polymerase chain reaction western blotting were used detect expression DANCR PTEN. MTT assay EdU performed cell proliferation cells with different treatment....

Medical image segmentation and classification tasks have become increasing accurate by employing deep neural networks. However, existing convolution networks models (CNNs) are challenging to achieve quite satisfactory results as medical objects backgrounds usually indistinguishable in spatial-domain images. In comparison, it is easier analyze complex frequency-domain images different object information retained frequency components. training CNNs the domain requires modification for network...

Cross-modal retrieval aims to retrieve relevant content of different modalities by giving a query another modality. The biggest difficulty is how bridge the heterogeneous gap between modalities. commonly-used methods tend focus on exploiting individual image-text pair and mining relations cross-modality data thereof, but ignore role multi-sample correlation. Moreover, more global, structural inter-pair knowledge contained training dataset will be under-used. To fully exploit graph-structured...

We are developing a computer-aided detection system to aid radiologists in diagnosing lung cancer thoracic computed tomographic (CT) images. The purpose of this study was improve the false-positive (FP) reduction stage our algorithm by and incorporating gradient field technique. This technique extracts 3D shape information from gray-scale values within volume interest. feature higher for spherical objects, lower elongated irregularly-shaped objects. A data set 55 thin CT scans 40 patients...

Deep neural networks trained by medical images with dense annotations have revealed favourable performance on accurate organ segmentation. The current supervised methods demand voxel-level which are not easily accessible due to the consuming of time and requirements specialized knowledge skills. In this paper, we propose a weakly method based recurrent residual convolutional network only image-level labels generate units take advantage contextual information successive slices spatial pooling...

A CAD system was developed to extract and analyze features from corresponding malignant benign lung nodules on temporal pairs of CT scans. The the current prior scans were automatically segmented using a 3-dimensional (3D) active contour model. Three-dimensional run length statistics (RLS) texture features, 3D morphological gray-level extracted each nodule. In addition, nodule profile (PROF) that describe gray level variation inside outside surface by estimating gradient magnitude values...

Automatic detection of chorioretinopathy plays an important role in clinical practice, but the a major central serous based on fundus photography images has rarely been studied, let alone distinguishing it from another exudative chorioretinopathy. Due to high degree similarity between two chorioretinopathies images, is difficult for latest automatic methods accurately distinguish them. In this study, we design deep neural network with branches different classification tasks, where first one...

The development scale of China’s sports industry continues to expand, and the speed accelerate, forming a complete industrial chain industry. At present, overall innovative renewal world’s stadium operations characteristics informatization, intelligence, integrated have been widely recognized. intelligent construction transformation smart stadiums become main direction future development. research objects were selected from 12 universities in Wuhan, China as examples, was empirical analyzed....

In recent years, knowledge distillation for semantic segmentation has been extensively studied in order to obtain satisfactory performance while reducing computational costs. Compared with natural images, targets medical images have fuzzy boundaries that are difficult determine, but current methods fail explicitly transfer boundary information, resulting poor discrimination compact models. Therefore, this paper, we propose two modules, namely boundary-guided deep supervision and output space...