Showing papers by "Yonsei University published in 2018"

Machine learning for molecular and materials science.

Abstract: Here we summarize recent progress in machine learning for the chemical sciences. We outline machine-learning techniques that are suitable for addressing research questions in this domain, as well as future directions for the field. We envisage a future in which the design, synthesis, characterization and application of molecules and materials is accelerated by artificial intelligence.

Single-cell RNA sequencing technologies and bioinformatics pipelines

Abstract: Rapid progress in the development of next-generation sequencing (NGS) technologies in recent years has provided many valuable insights into complex biological systems, ranging from cancer genomics to diverse microbial communities. NGS-based technologies for genomics, transcriptomics, and epigenomics are now increasingly focused on the characterization of individual cells. These single-cell analyses will allow researchers to uncover new and potentially unexpected biological discoveries relative to traditional profiling methods that assess bulk populations. Single-cell RNA sequencing (scRNA-seq), for example, can reveal complex and rare cell populations, uncover regulatory relationships between genes, and track the trajectories of distinct cell lineages in development. In this review, we will focus on technical challenges in single-cell isolation and library preparation and on computational analysis pipelines available for analyzing scRNA-seq data. Further technical improvements at the level of molecular and cell biology and in available bioinformatics tools will greatly facilitate both the basic science and medical applications of these sequencing technologies.

Deep Learning Techniques for Automatic MRI Cardiac Multi-Structures Segmentation and Diagnosis: Is the Problem Solved?

Abstract: Delineation of the left ventricular cavity, myocardium, and right ventricle from cardiac magnetic resonance images (multi-slice 2-D cine MRI) is a common clinical task to establish diagnosis. The automation of the corresponding tasks has thus been the subject of intense research over the past decades. In this paper, we introduce the “Automatic Cardiac Diagnosis Challenge” dataset (ACDC), the largest publicly available and fully annotated dataset for the purpose of cardiac MRI (CMR) assessment. The dataset contains data from 150 multi-equipments CMRI recordings with reference measurements and classification from two medical experts. The overarching objective of this paper is to measure how far state-of-the-art deep learning methods can go at assessing CMRI, i.e., segmenting the myocardium and the two ventricles as well as classifying pathologies. In the wake of the 2017 MICCAI-ACDC challenge, we report results from deep learning methods provided by nine research groups for the segmentation task and four groups for the classification task. Results show that the best methods faithfully reproduce the expert analysis, leading to a mean value of 0.97 correlation score for the automatic extraction of clinical indices and an accuracy of 0.96 for automatic diagnosis. These results clearly open the door to highly accurate and fully automatic analysis of cardiac CMRI. We also identify scenarios for which deep learning methods are still failing. Both the dataset and detailed results are publicly available online, while the platform will remain open for new submissions.

TRRUST v2: an expanded reference database of human and mouse transcriptional regulatory interactions.

Abstract: Transcription factors (TFs) are major trans-acting factors in transcriptional regulation. Therefore, elucidating TF-target interactions is a key step toward understanding the regulatory circuitry underlying complex traits such as human diseases. We previously published a reference TF-target interaction database for humans-TRRUST (Transcriptional Regulatory Relationships Unraveled by Sentence-based Text mining)-which was constructed using sentence-based text mining, followed by manual curation. Here, we present TRRUST v2 (www.grnpedia.org/trrust) with a significant improvement from the previous version, including a significantly increased size of the database consisting of 8444 regulatory interactions for 800 TFs in humans. More importantly, TRRUST v2 also contains a database for TF-target interactions in mice, including 6552 TF-target interactions for 828 mouse TFs. TRRUST v2 is also substantially more comprehensive and less biased than other TF-target interaction databases. We also improved the web interface, which now enables prioritization of key TFs for a physiological condition depicted by a set of user-input transcriptional responsive genes. With the significant expansion in the database size and inclusion of the new web tool for TF prioritization, we believe that TRRUST v2 will be a versatile database for the study of the transcriptional regulation involved in human diseases.

Identification of distinct nanoparticles and subsets of extracellular vesicles by asymmetric flow field-flow fractionation.

Abstract: The heterogeneity of exosomal populations has hindered our understanding of their biogenesis, molecular composition, biodistribution and functions. By employing asymmetric flow field-flow fractionation (AF4), we identified two exosome subpopulations (large exosome vesicles, Exo-L, 90–120 nm; small exosome vesicles, Exo-S, 60–80 nm) and discovered an abundant population of non-membranous nanoparticles termed ‘exomeres’ (~35 nm). Exomere proteomic profiling revealed an enrichment in metabolic enzymes and hypoxia, microtubule and coagulation proteins as well as specific pathways, such as glycolysis and mTOR signalling. Exo-S and Exo-L contained proteins involved in endosomal function and secretion pathways, and mitotic spindle and IL-2/STAT5 signalling pathways, respectively. Exo-S, Exo-L and exomeres each had unique N-glycosylation, protein, lipid, DNA and RNA profiles and biophysical properties. These three nanoparticle subsets demonstrated diverse organ biodistribution patterns, suggesting distinct biological functions. This study demonstrates that AF4 can serve as an improved analytical tool for isolating extracellular vesicles and addressing the complexities of heterogeneous nanoparticle subpopulations.

Tokyo Guidelines 2018: diagnostic criteria and severity grading of acute cholecystitis (with videos).

Abstract: Although the diagnostic and severity grading criteria on the 2013 Tokyo Guidelines (TG13) are used worldwide as the primary standard for management of acute cholangitis (AC), they need to be validated through implementation and assessment in actual clinical practice. Here, we conduct a systematic review of the literature to validate the TG13 diagnostic and severity grading criteria for AC and propose TG18 criteria. While there is little evidence evaluating the TG13 criteria, they were validated through a large-scale case series study in Japan and Taiwan. Analyzing big data from this study confirmed that the diagnostic rate of AC based on the TG13 diagnostic criteria was higher than that based on the TG07 criteria, and that 30-day mortality in patients with a higher severity based on the TG13 severity grading criteria was significantly higher. Furthermore, a comparison of patients treated with early or urgent biliary drainage versus patients not treated this way showed no difference in 30-day mortality among patients with Grade I or Grade III AC, but significantly lower 30-day mortality in patients with Grade II AC who were treated with early or urgent biliary drainage. This suggests that the TG13 severity grading criteria can be used to identify Grade II patients whose prognoses may be improved through biliary drainage. The TG13 severity grading criteria may therefore be useful as an indicator for biliary drainage as well as a predictive factor when assessing the patient's prognosis. The TG13 diagnostic and severity grading criteria for AC can provide results quickly, are minimally invasive for the patients, and are inexpensive. We recommend that the TG13 criteria be adopted in the TG18 guidelines and used as standard practice in the clinical setting. Free full articles and mobile app of TG18 are available at: http://www.jshbps.jp/modules/en/index.php?content_id=47. Related clinical questions and references are also included.

El Niño–Southern Oscillation complexity

Abstract: El Nino events are characterized by surface warming of the tropical Pacific Ocean and weakening of equatorial trade winds that occur every few years Such conditions are accompanied by changes in atmospheric and oceanic circulation, affecting global climate, marine and terrestrial ecosystems, fisheries and human activities The alternation of warm El Nino and cold La Nina conditions, referred to as the El Nino–Southern Oscillation (ENSO), represents the strongest year-to-year fluctuation of the global climate system Here we provide a synopsis of our current understanding of the spatio-temporal complexity of this important climate mode and its influence on the Earth system

Dabrafenib and Trametinib Treatment in Patients With Locally Advanced or Metastatic BRAF V600-Mutant Anaplastic Thyroid Cancer

Abstract: Purpose We report the efficacy and safety of dabrafenib (BRAF inhibitor) and trametinib (MEK inhibitor) combination therapy in BRAF V600E-mutated anaplastic thyroid cancer, a rare, aggressive, and highly lethal malignancy with poor patient outcomes and no systemic therapies with clinical benefit. Methods In this phase II, open-label trial, patients with predefined BRAF V600E-mutated malignancies received dabrafenib 150 mg twice daily and trametinib 2 mg once daily until unacceptable toxicity, disease progression, or death. The primary end point was investigator-assessed overall response rate. Secondary end points included duration of response, progression-free survival, overall survival, and safety. Results Sixteen patients with BRAF V600E-mutated anaplastic thyroid cancer were evaluable (median follow-up, 47 weeks; range, 4 to 120 weeks). All patients had received prior radiation treatment and/or surgery, and six had received prior systemic therapy. The confirmed overall response rate was 69% (11 of 16; 95% CI, 41% to 89%), with seven ongoing responses. Median duration of response, progression-free survival, and overall survival were not reached as a result of a lack of events, with 12-month estimates of 90%, 79%, and 80%, respectively. The safety population was composed of 100 patients who were enrolled with seven rare tumor histologies. Common adverse events were fatigue (38%), pyrexia (37%), and nausea (35%). No new safety signals were detected. Conclusion Dabrafenib plus trametinib is the first regimen demonstrated to have robust clinical activity in BRAF V600E-mutated anaplastic thyroid cancer and was well tolerated. These findings represent a meaningful therapeutic advance for this orphan disease.

Methodologic Guide for Evaluating Clinical Performance and Effect of Artificial Intelligence Technology for Medical Diagnosis and Prediction

Abstract: In this article, we review some of the key methodologic points that should be considered with regard to clinical evaluation of artificial intelligence tools for use in medical diagnosis and prediction.

A Survey on Hybrid Beamforming Techniques in 5G: Architecture and System Model Perspectives

Abstract: The increasing wireless data traffic demands have driven the need to explore suitable spectrum regions for meeting the projected requirements. In the light of this, millimeter wave (mmWave) communication has received considerable attention from the research community. Typically, in fifth generation (5G) wireless networks, mmWave massive multiple-input multiple-output (MIMO) communications is realized by the hybrid transceivers which combine high dimensional analog phase shifters and power amplifiers with lower-dimensional digital signal processing units. This hybrid beamforming design reduces the cost and power consumption which is aligned with an energy-efficient design vision of 5G. In this paper, we track the progress in hybrid beamforming for massive MIMO communications in the context of system models of the hybrid transceivers’ structures, the digital and analog beamforming matrices with the possible antenna configuration scenarios and the hybrid beamforming in heterogeneous wireless networks. We extend the scope of the discussion by including resource management issues in hybrid beamforming. We explore the suitability of hybrid beamforming methods, both, existing and proposed till first quarter of 2017, and identify the exciting future challenges in this domain.

Deep Video Super-Resolution Network Using Dynamic Upsampling Filters Without Explicit Motion Compensation

Abstract: Video super-resolution (VSR) has become even more important recently to provide high resolution (HR) contents for ultra high definition displays. While many deep learning based VSR methods have been proposed, most of them rely heavily on the accuracy of motion estimation and compensation. We introduce a fundamentally different framework for VSR in this paper. We propose a novel end-to-end deep neural network that generates dynamic upsampling filters and a residual image, which are computed depending on the local spatio-temporal neighborhood of each pixel to avoid explicit motion compensation. With our approach, an HR image is reconstructed directly from the input image using the dynamic upsampling filters, and the fine details are added through the computed residual. Our network with the help of a new data augmentation technique can generate much sharper HR videos with temporal consistency, compared with the previous methods. We also provide analysis of our network through extensive experiments to show how the network deals with motions implicitly.

The Global Precipitation Climatology Project (GPCP) Monthly Analysis (New Version 2.3) and a Review of 2017 Global Precipitation

Abstract: The new Version 2.3 of the Global Precipitation Climatology Project (GPCP) Monthly analysis is described in terms of changes made to improve the homogeneity of the product, especially after 2002. These changes include corrections to cross-calibration of satellite data inputs and updates to the gauge analysis. Over-ocean changes starting in 2003 resulted in an overall precipitation increase of 1.8% after 2009. Updating the gauge analysis to its final, high-quality version increases the global land total by 1.8% for the post-2002 period. These changes correct a small, incorrect dip in the estimated global precipitation over the last decade given by the earlier Version 2.2. The GPCP analysis is also used to describe global precipitation in 2017. The general La Nina pattern for 2017 is noted and the evolution from the early 2016 El Nino pattern is described. The 2017 global value is one of the highest for the 1979–2017 period, exceeded only by 2016 and 1998 (both El Nino years), and reinforces the small positive trend. Results for 2017 also reinforce significant trends in precipitation intensity (on a monthly scale) in the tropics. These results for 2017 indicate the value of the GPCP analysis, in addition to research, for climate monitoring.

Multi-Ethnic Genome-wide Association Study for Atrial Fibrillation

Abstract: Atrial fibrillation (AF) affects more than 33 million individuals worldwide1 and has a complex heritability2. We conducted the largest meta-analysis of genome-wide association studies (GWAS) for AF to date, consisting of more than half a million individuals, including 65,446 with AF. In total, we identified 97 loci significantly associated with AF, including 67 that were novel in a combined-ancestry analysis, and 3 that were novel in a European-specific analysis. We sought to identify AF-associated genes at the GWAS loci by performing RNA-sequencing and expression quantitative trait locus analyses in 101 left atrial samples, the most relevant tissue for AF. We also performed transcriptome-wide analyses that identified 57 AF-associated genes, 42 of which overlap with GWAS loci. The identified loci implicate genes enriched within cardiac developmental, electrophysiological, contractile and structural pathways. These results extend our understanding of the biological pathways underlying AF and may facilitate the development of therapeutics for AF.

CNS Response to Osimertinib Versus Standard Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitors in Patients With Untreated EGFR-Mutated Advanced Non–Small-Cell Lung Cancer

Abstract: Purpose We report CNS efficacy of osimertinib versus standard epidermal growth factor receptor ( EGFR) tyrosine kinase inhibitors (TKIs) in patients with untreated EGFR-mutated advanced non-small-cell lung cancer from the phase III FLAURA study. Patients and Methods Patients (N = 556) were randomly assigned to osimertinib or standard EGFR-TKIs (gefitinib or erlotinib); brain scans were not mandated unless clinically indicated. Patients with asymptomatic or stable CNS metastases were included. In patients with symptomatic CNS metastases, neurologic status was required to be stable for ≥ 2 weeks after completion of definitive therapy and corticosteroids. A preplanned subgroup analysis with CNS progression-free survival as primary objective was conducted in patients with measurable and/or nonmeasurable CNS lesions on baseline brain scan by blinded independent central neuroradiologic review. The CNS evaluable-for-response set included patients with ≥ one measurable CNS lesion. Results Of 200 patients with available brain scans at baseline, 128 (osimertinib, n = 61; standard EGFR-TKIs, n = 67) had measurable and/or nonmeasurable CNS lesions, including 41 patients (osimertinib, n = 22; standard EGFR-TKIs, n = 19) with ≥ one measurable CNS lesion. Median CNS progression-free survival in patients with measurable and/or nonmeasurable CNS lesions was not reached with osimertinib (95% CI, 16.5 months to not calculable) and 13.9 months (95% CI, 8.3 months to not calculable) with standard EGFR-TKIs (hazard ratio, 0.48; 95% CI, 0.26 to 0.86; P = .014 [nominally statistically significant]). CNS objective response rates were 91% and 68% in patients with ≥ one measurable CNS lesion (odds ratio, 4.6; 95% CI, 0.9 to 34.9; P = .066) and 66% and 43% in patients with measurable and/or nonmeasurable CNS lesions (odds ratio, 2.5; 95% CI, 1.2 to 5.2; P = .011) treated with osimertinib and standard EGFR-TKIs, respectively. Probability of experiencing a CNS progression event was consistently lower with osimertinib versus standard EGFR-TKIs. Conclusion Osimertinib has CNS efficacy in patients with untreated EGFR-mutated non-small-cell lung cancer. These results suggest a reduced risk of CNS progression with osimertinib versus standard EGFR-TKIs.

SIRveNIB: Selective Internal Radiation Therapy Versus Sorafenib in Asia-Pacific Patients With Hepatocellular Carcinoma

Abstract: Purpose Selective internal radiation therapy or radioembolization (RE) shows efficacy in unresectable hepatocellular carcinoma (HCC) limited to the liver. This study compared the safety and efficacy of RE and sorafenib in patients with locally advanced HCC. Patients and Methods SIRveNIB (selective internal radiation therapy v sorafenib), an open-label, investigator-initiated, phase III trial, compared yttrium-90 (90Y) resin microspheres RE with sorafenib 800 mg/d in patients with locally advanced HCC in a two-tailed study designed for superiority/detriment. Patients were randomly assigned 1:1 and stratified by center and presence of portal vein thrombosis. Primary end point was overall survival (OS). Efficacy analyses were performed in the intention-to-treat population and safety analyses in the treated population. Results A total of 360 patients were randomly assigned (RE, 182; sorafenib, 178) from 11 countries in the Asia-Pacific region. In the RE and sorafenib groups, 28.6% and 9.0%, respectively, failed to receive assigned therapy without significant cross-over to either group. Median OS was 8.8 and 10.0 months with RE and sorafenib, respectively (hazard ratio, 1.1; 95% CI, 0.9 to 1.4; P = .36). A total of 1,468 treatment-emergent adverse events (AEs) were reported (RE, 437; sorafenib, 1,031). Significantly fewer patients in the RE than sorafenib group had grade ≥ 3 AEs (36 of 130 [27.7%]) v 82 of 162 [50.6%]; P < .001). The most common grade ≥ 3 AEs were ascites (five of 130 [3.8%] v four of 162 [2.5%] patients), abdominal pain (three [2.3%] v two [1.2%] patients), anemia (zero v four [2.5%] patients), and radiation hepatitis (two [1.5%] v zero [0%] patients). Fewer patients in the RE group (27 of 130 [20.8%]) than in the sorafenib group (57 of 162 [35.2%]) had serious AEs. Conclusion In patients with locally advanced HCC, OS did not differ significantly between RE and sorafenib. The improved toxicity profile of RE may inform treatment choice in selected patients.

Tokyo Guidelines 2018: flowchart for the management of acute cholecystitis.

Abstract: We propose a new flowchart for the treatment of acute cholecystitis (AC) in the Tokyo Guidelines 2018 (TG18). Grade III AC was not indicated for straightforward laparoscopic cholecystectomy (Lap-C). Following analysis of subsequent clinical investigations and drawing on Big Data in particular, TG18 proposes that some Grade III AC can be treated by Lap-C when performed at advanced centers with specialized surgeons experienced in this procedure and for patients that satisfy certain strict criteria. For Grade I, TG18 recommends early Lap-C if the patients meet the criteria of Charlson comorbidity index (CCI) ≤5 and American Society of Anesthesiologists physical status classification (ASA-PS) ≤2. For Grade II AC, if patients meet the criteria of CCI ≤5 and ASA-PS ≤2, TG18 recommends early Lap-C performed by experienced surgeons; and if not, after medical treatment and/or gallbladder drainage, Lap-C would be indicated. TG18 proposes that Lap-C is indicated in Grade III patients with strict criteria. These are that the patients have favorable organ system failure, and negative predictive factors, who meet the criteria of CCI ≤3 and ASA-PS ≤2 and who are being treated at an advanced center (where experienced surgeons practice). If the patient is not considered suitable for early surgery, TG18 recommends early/urgent biliary drainage followed by delayed Lap-C once the patient's overall condition has improved. Free full articles and mobile app of TG18 are available at: http://www.jshbps.jp/modules/en/index.php?content_id=47. Related clinical questions and references are also included.

Rhizosphere microbiome structure alters to enable wilt resistance in tomato

Abstract: Tomato variety Hawaii 7996 is resistant to the soil-borne pathogen Ralstonia solanacearum, whereas the Moneymaker variety is susceptible to the pathogen. To evaluate whether plant-associated microorganisms have a role in disease resistance, we analyzed the rhizosphere microbiomes of both varieties in a mesocosm experiment. Microbiome structures differed between the two cultivars. Transplantation of rhizosphere microbiota from resistant plants suppressed disease symptoms in susceptible plants. Comparative analyses of rhizosphere metagenomes from resistant and susceptible plants enabled the identification and assembly of a flavobacterial genome that was far more abundant in the resistant plant rhizosphere microbiome than in that of the susceptible plant. We cultivated this flavobacterium, named TRM1, and found that it could suppress R. solanacearum-disease development in a susceptible plant in pot experiments. Our findings reveal a role for native microbiota in protecting plants from microbial pathogens, and our approach charts a path toward the development of probiotics to ameliorate plant diseases.

Human glioblastoma arises from subventricular zone cells with low-level driver mutations

Abstract: Glioblastoma (GBM) is a devastating and incurable brain tumour, with a median overall survival of fifteen months1,2. Identifying the cell of origin that harbours mutations that drive GBM could provide a fundamental basis for understanding disease progression and developing new treatments. Given that the accumulation of somatic mutations has been implicated in gliomagenesis, studies have suggested that neural stem cells (NSCs), with their self-renewal and proliferative capacities, in the subventricular zone (SVZ) of the adult human brain may be the cells from which GBM originates3–5. However, there is a lack of direct genetic evidence from human patients with GBM4,6–10. Here we describe direct molecular genetic evidence from patient brain tissue and genome-edited mouse models that show astrocyte-like NSCs in the SVZ to be the cell of origin that contains the driver mutations of human GBM. First, we performed deep sequencing of triple-matched tissues, consisting of (i) normal SVZ tissue away from the tumour mass, (ii) tumour tissue, and (iii) normal cortical tissue (or blood), from 28 patients with isocitrate dehydrogenase (IDH) wild-type GBM or other types of brain tumour. We found that normal SVZ tissue away from the tumour in 56.3% of patients with wild-type IDH GBM contained low-level GBM driver mutations (down to approximately 1% of the mutational burden) that were observed at high levels in their matching tumours. Moreover, by single-cell sequencing and laser microdissection analysis of patient brain tissue and genome editing of a mouse model, we found that astrocyte-like NSCs that carry driver mutations migrate from the SVZ and lead to the development of high-grade malignant gliomas in distant brain regions. Together, our results show that NSCs in human SVZ tissue are the cells of origin that contain the driver mutations of GBM.

Radial-artery or saphenous-vein grafts in coronary-artery bypass surgery

Abstract: Background The use of radial-artery grafts for coronary-artery bypass grafting (CABG) may result in better postoperative outcomes than the use of saphenous-vein grafts. However, randomized, controlled trials comparing radial-artery grafts and saphenous-vein grafts have been individually underpowered to detect differences in clinical outcomes. We performed a patient-level combined analysis of randomized, controlled trials to compare radial-artery grafts and saphenous-vein grafts for CABG. Methods Six trials were identified. The primary outcome was a composite of death, myocardial infarction, or repeat revascularization. The secondary outcome was graft patency on follow-up angiography. Mixed-effects Cox regression models were used to estimate the treatment effect on the outcomes. Results A total of 1036 patients were included in the analysis (534 patients with radial-artery grafts and 502 patients with saphenous-vein grafts). After a mean (±SD) follow-up time of 60±30 months, the incidence of adver...

Fast Video Object Segmentation by Reference-Guided Mask Propagation

Abstract: We present an efficient method for the semi-supervised video object segmentation. Our method achieves accuracy competitive with state-of-the-art methods while running in a fraction of time compared to others. To this end, we propose a deep Siamese encoder-decoder network that is designed to take advantage of mask propagation and object detection while avoiding the weaknesses of both approaches. Our network, learned through a two-stage training process that exploits both synthetic and real data, works robustly without any online learning or post-processing. We validate our method on four benchmark sets that cover single and multiple object segmentation. On all the benchmark sets, our method shows comparable accuracy while having the order of magnitude faster runtime. We also provide extensive ablation and add-on studies to analyze and evaluate our framework.

Asia-Pacific Working Party on Non-alcoholic Fatty Liver Disease guidelines 2017-Part 1: Definition, risk factors and assessment.

Abstract: Asia–Pacific Working Party on Non-alcoholic Fatty Liver Disease guidelines 2017—Part 1: Definition, risk factors and assessment Vincent Wai-Sun Wong,* Wah-Kheong Chan, Shiv Chitturi, Yogesh Chawla, Yock Young Dan,** Ajay Duseja, Jiangao Fan, Khean-Lee Goh, Masahide Hamaguchi, Etsuko Hashimoto, Seung Up Kim, Laurentius Adrianto Lesmana,*** Yu-Cheng Lin, Chun-Jen Liu, Yen-Hsuan Ni, Jose Sollano, Simon Kin-Hung Wong, Grace Lai-Hung Wong,* Henry Lik-Yuen Chan* and Geoff Farrell *Department ofMedicine and Therapeutics, State Key Laboratory of Digestive Disease andDepartment of Surgery, Department of Surgery, The Chinese University of Hong Kong, Shatin, Hong Kong; Department of Medicine, University of Malaya, Kuala Lumpur, Malaysia; Gastroenterology and Hepatology Unit, The Canberra Hospital, Canberra, Australian Capital Territory, Australia; Department of Hepatology, Postgraduate Institute of Medical Education and Research, Chandigarh, India; **Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, Singapore; Xin Hua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China; Department of Diabetology, Kameoka Municipal Hospital, Kameoka and Departments of InternalMedicine andGastroenterology, TokyoWomen’sMedical University, Tokyo, Japan; Department of InternalMedicine, Institute of Gastroenterology, Yonsei University College ofMedicine, Seoul, Korea; ***Digestive Disease andGI Oncology Centre,Medistra Hospital, Jakarta, Indonesia; Hepatitis Research Center, National Taiwan University, and Department of Internal Medicine, Hepatitis Research Center and Graduate Institute of Clinical Medicine, National Taiwan University College of Medicine and Hospital, Taipei, Taiwan; and University of Santo Tomas, Manila, The Philippines

KIKI-net: cross-domain convolutional neural networks for reconstructing undersampled magnetic resonance images.

Abstract: Purpose To demonstrate accurate MR image reconstruction from undersampled k-space data using cross-domain convolutional neural networks (CNNs) METHODS: Cross-domain CNNs consist of 3 components: (1) a deep CNN operating on the k-space (KCNN), (2) a deep CNN operating on an image domain (ICNN), and (3) an interleaved data consistency operations. These components are alternately applied, and each CNN is trained to minimize the loss between the reconstructed and corresponding fully sampled k-spaces. The final reconstructed image is obtained by forward-propagating the undersampled k-space data through the entire network. Results Performances of K-net (KCNN with inverse Fourier transform), I-net (ICNN with interleaved data consistency), and various combinations of the 2 different networks were tested. The test results indicated that K-net and I-net have different advantages/disadvantages in terms of tissue-structure restoration. Consequently, the combination of K-net and I-net is superior to single-domain CNNs. Three MR data sets, the T2 fluid-attenuated inversion recovery (T2 FLAIR) set from the Alzheimer's Disease Neuroimaging Initiative and 2 data sets acquired at our local institute (T2 FLAIR and T1 weighted), were used to evaluate the performance of 7 conventional reconstruction algorithms and the proposed cross-domain CNNs, which hereafter is referred to as KIKI-net. KIKI-net outperforms conventional algorithms with mean improvements of 2.29 dB in peak SNR and 0.031 in structure similarity. Conclusion KIKI-net exhibits superior performance over state-of-the-art conventional algorithms in terms of restoring tissue structures and removing aliasing artifacts. The results demonstrate that KIKI-net is applicable up to a reduction factor of 3 to 4 based on variable-density Cartesian undersampling.