Free-form tumor synthesis in computed tomography images via richer generative adversarial network

Abstract

The insufficiency of annotated medical imaging scans for cancer makes it challenging to train and validate data-hungry deep learning models in precision oncology. We propose a new richer generative adversarial network for free-form 3D tumor/lesion synthesis in computed tomography (CT) images. The network is composed of a new richer convolutional feature enhanced dilated-gated generator (RicherDG) and a hybrid loss function. The RicherDG has dilated-gated convolution layers to enable tumor-painting and to enlarge perceptive fields; and it has a novel richer convolutional feature association branch to recover multi-scale convolutional features especially from uncertain boundaries between tumor and surrounding healthy tissues. The hybrid loss function, which consists of a diverse range of losses, is designed to aggregate complementary information to improve optimization. We perform a comprehensive evaluation of the synthesis results on a wide range of public CT image datasets covering the liver, kidney tumors, and lung nodules. The qualitative and quantitative evaluations and ablation study demonstrated improved synthesizing results over advanced tumor synthesis methods.

DrugHybrid_BS: Using Hybrid Feature Combined With Bagging-SVM to Predict Potentially Druggable Proteins

“…Bagging is one of the common ensemble learning models ( Dudoit and Fridlyand, 2003 ; Jin et al, 2019 ; Jin et al, 2021 ; Wu and Yu, 2021 ). The ensemble learning model uses a series of weak learners (also called basic models) for learning and integrates the results of each weak learner to obtain a better learning effect than individual learners.…”

Genomic Variation Prediction: A Summary From Different Views

“…Pathogenic synonymous mutations SilVA (Buske et al, 2013) Random forest DDIG-SN (Livingstone et al, 2017) Support vector machine regSNPs-splicing (Zhang et al, 2017a) Random forest Syntool (Zhang et al, 2017b) -TraP (Gelfman et al, 2017) Random forest Genome sequencing CADD (Kircher et al, 2014) Support vector machine MutationTaster2 (Cooper, 2014) Naive Bayes Mut-Pred (Li et al, 2009) Random forest PolyPhen-2 (Adzhubei et al, 2010) Naive Bayes PON-P2 (Niroula et al, 2015) Random forest VEST (Carter et al, 2013) Random forest Deep mining of structural variation information DeepBind (Alipanahi et al, 2015) deep learning DeepVariant (Angermueller et al, 2017) deep neural networks DeepCpG (Poplin et al, 2018) deep 2019; Liu et al, 2020a;Jin et al, 2021;Yin et al, 2021). The published methods include CADD (Kircher et al, 2014), MutationTaster2 (Cooper, 2014),Mut-Pred (Li et al, 2009),PolyPhen-2 (Adzhubei et al, 2010),PON-P2 (Niroula et al, 2015) and VEST (Carter et al, 2013).…”

“…Among them, the use of machine learning methods to predict pathogenic synonymous mutations is still in the preliminary stage. The main problems that remain to be solved include: 1) positive sample data is scarce and standard negative sample data is lacking ( Zhang et al, 2020b ); 2) feature representation ability is weak and not easy to promote ( Buske et al, 2013 ; Wei et al, 2018 ; Xiong et al, 2018 ; Jin et al, 2019 ; Shen et al, 2019 ; Su et al, 2019 ; Wei et al, 2019 ; Yang et al, 2020a ; Zhang et al, 2020c ; Peng et al, 2020 ; Su et al, 2020 ; Teng et al, 2020 ; Chu et al, 2021a ; Cheng et al, 2021b ; Chu et al, 2021b ; Jin et al, 2021 ; Su et al, 2021 ); and 3) the prediction performances of existing methods need to be improved, and the results of different methods have a low degree of coincidence ( Cheng et al, 2019 ). The methods reviewed in this article aim to solve these problems.…”

“…A primary task of sequencing research is how to analyze the sequence data, especially the variation information that it contains. Prediction methods can effectively solve this problem ( Castrense et al, 2019 ; Gang et al, 2019 ; Liu et al, 2020a ; Jin et al, 2021 ; Yin et al, 2021 ). The published methods include CADD ( Kircher et al, 2014 ), MutationTaster2 ( Cooper, 2014 ),Mut-Pred ( Li et al, 2009 ),PolyPhen-2 ( Adzhubei et al, 2010 ),PON-P2 ( Niroula et al, 2015 ) and VEST ( Carter et al, 2013 ).…”

RF‐GANs: A Method to Synthesize Retinal Fundus Images Based on Generative Adversarial Network

“…Jiang et al [ 26 ] extended CGAN by employing dual generators and dual discriminators that introduced a dynamic communication mechanism to improve CGAN to synthesize lung computed tomography (CT) images, then combined the generated lung CT images with nonpulmonary CT to get COVID-19 chest X-ray. Jin et al [ 27 ] employed a new, richer convolutional feature enhanced dilated-gated generator (RicherDG) to synthesize 3D tumor lesions in CT images.…”