Deep learning-based age estimation from chest X-rays indicates cardiovascular prognosis

BACKGROUND

In recent years, there has been considerable research on the use of artificial intelligence to estimate age and disease status from medical images. However, age estimation from chest X-ray (CXR) images has not been well studied and the clinical significance of estimated age has not been fully determined.

METHODS

To address this, we trained a deep neural network (DNN) model using more than 100,000 CXRs to estimate the patients' age solely from CXRs. We applied our DNN to CXRs of 1562 consecutive hospitalized heart failure patients, and 3586 patients admitted to the intensive care unit with cardiovascular disease.

RESULTS

The DNN's estimated age (X-ray age) showed a strong significant correlation with chronological age on the hold-out test data and independent test data. Elevated X-ray age is associated with worse clinical outcomes (heart failure readmission and all-cause death) for heart failure. Additionally, elevated X-ray age was associated with a worse prognosis in 3586 patients admitted to the intensive care unit with cardiovascular disease.

CONCLUSIONS

Our results suggest that X-ray age can serve as a useful indicator of cardiovascular abnormalities, which will help clinicians to predict, prevent and manage cardiovascular diseases.

The Effectiveness of a Deep Learning Model to Detect Left Ventricular Systolic Dysfunction from Electrocardiograms

Deep learning models can be applied to electrocardiograms (ECGs) to detect left ventricular (LV) dysfunction. We hypothesized that applying a deep learning model may improve the diagnostic accuracy of cardiologists in predicting LV dysfunction from ECGs. We acquired 37,103 paired ECG and echocardiography data records of patients who underwent echocardiography between January 2015 and December 2019. We trained a convolutional neural network to identify the data records of patients with LV dysfunction (ejection fraction < 40%) using a dataset of 23,801 ECGs. When tested on an independent set of 7,196 ECGs, we found the area under the receiver operating characteristic curve was 0.945 (95% confidence interval: 0.936-0.954). When 7 cardiologists interpreted 50 randomly selected ECGs from the test dataset of 7,196 ECGs, their accuracy for predicting LV dysfunction was 78.0% ± 6.0%. By referring to the model's output, the cardiologist accuracy improved to 88.0% ± 3.7%, which indicates that model support significantly improved the cardiologist diagnostic accuracy (P = 0.02). A sensitivity map demonstrated that the model focused on the QRS complex when detecting LV dysfunction on ECGs. We developed a deep learning model that can detect LV dysfunction on ECGs with high accuracy. Furthermore, we demonstrated that support from a deep learning model can help cardiologists to identify LV dysfunction on ECGs.

Deep Learning Algorithm to Detect Cardiac Sarcoidosis From Echocardiographic Movies

BACKGROUND

Because the early diagnosis of subclinical cardiac sarcoidosis (CS) remains difficult, we developed a deep learning algorithm to distinguish CS patients from healthy subjects using echocardiographic movies.Methods and Results:Among the patients who underwent echocardiography from January 2015 to December 2019, we chose 151 echocardiographic movies from 50 CS patients and 151 from 149 healthy subjects. We trained two 3D convolutional neural networks (3D-CNN) to identify CS patients using a dataset of 212 echocardiographic movies with and without a transfer learning method (Pretrained algorithm and Non-pretrained algorithm). On an independent set of 41 echocardiographic movies, the area under the receiver-operating characteristic curve (AUC) of the Pretrained algorithm was greater than that of Non-pretrained algorithm (0.842, 95% confidence interval (CI): 0.722-0.962 vs. 0.724, 95% CI: 0.566-0.882, P=0.253). The AUC from the interpretation of the same set of 41 echocardiographic movies by 5 cardiologists was not significantly different from that of the Pretrained algorithm (0.855, 95% CI: 0.735-0.975 vs. 0.842, 95% CI: 0.722-0.962, P=0.885). A sensitivity map demonstrated that the Pretrained algorithm focused on the area of the mitral valve.

CONCLUSIONS

A 3D-CNN with a transfer learning method may be a promising tool for detecting CS using an echocardiographic movie.

Population-specific and trans-ancestry genome-wide analyses identify distinct and shared genetic risk loci for coronary artery disease

To elucidate the genetics of coronary artery disease (CAD) in the Japanese population, we conducted a large-scale genome-wide association study of 168,228 individuals of Japanese ancestry (25,892 cases and 142,336 controls) with genotype imputation using a newly developed reference panel of Japanese haplotypes including 1,781 CAD cases and 2,636 controls. We detected eight new susceptibility loci and Japanese-specific rare variants contributing to disease severity and increased cardiovascular mortality. We then conducted a trans-ancestry meta-analysis and discovered 35 additional new loci. Using the meta-analysis results, we derived a polygenic risk score (PRS) for CAD, which outperformed those derived from either Japanese or European genome-wide association studies. The PRS prioritized risk factors among various clinical parameters and segregated individuals with increased risk of long-term cardiovascular mortality. Our data improve the clinical characterization of CAD genetics and suggest the utility of trans-ancestry meta-analysis for PRS derivation in non-European populations.

Erratum: Diagnosing Heart Failure from Chest X-Ray Images Using Deep Learning

An error appeared in the article entitled "Diagnosing Heart Failure from Chest X-Ray Images Using Deep Learning" by Takuya Matsumoto, Satoshi Kodera, Hiroki Shinohara, Hirotaka Ieki, Toshihiro Yamaguchi, Yasutomi Higashikuni, Arihiro Kiyosue, Kaoru Ito, Jiro Ando, Eiki Takimoto, Hiroshi Akazawa, Hiroyuki Morita, Issei Komuro (Vol. 61, No. 4, 781-786, 2020). The Figure 5on page 784 should be replaced by the following figure.

Diagnosing Heart Failure from Chest X-Ray Images Using Deep Learning

The development of deep learning technology has enabled machines to achieve high-level accuracy in interpreting medical images. While many previous studies have examined the detection of pulmonary nodules in chest X-rays using deep learning, the application of this technology to heart failure remains rare. In this paper, we investigated the performance of a deep learning algorithm in terms of diagnosing heart failure using images obtained from chest X-rays. We used 952 chest X-ray images from a labeled database published by the National Institutes of Health. Two cardiologists verified and relabeled a total of 260 "normal" and 378 "heart failure" images, with the remainder being discarded because they had been incorrectly labeled. Data augmentation and transfer learning were used to obtain an accuracy of 82% in diagnosing heart failure using the chest X-ray images. Furthermore, heatmap imaging allowed us to visualize decisions made by the machine. Deep learning can thus help support the diagnosis of heart failure using chest X-ray images.

Transethnic Meta-Analysis of Genome-Wide Association Studies Identifies Three New Loci and Characterizes Population-Specific Differences for Coronary Artery Disease

BACKGROUND

Genome-wide association studies provided many biological insights into coronary artery disease (CAD), but these studies were mainly performed in Europeans. Genome-wide association studies in diverse populations have the potential to advance our understanding of CAD.

METHODS

We conducted 2 genome-wide association studies for CAD in the Japanese population, which included 12 494 cases and 28 879 controls and 2808 cases and 7261 controls, respectively. Then, we performed transethnic meta-analysis using the results of the coronary artery disease genome-wide replication and meta-analysis plus the coronary artery disease 1000 Genomes meta-analysis with UK Biobank. We then explored the pathophysiological significance of these novel loci and examined the differences in CAD-susceptibility loci between Japanese and Europeans.

RESULTS

We identified 3 new loci on chromosome 1q21 (CTSS), 10q26 (WDR11-FGFR2), and 11q22 (RDX-FDX1). Quantitative trait locus analyses suggested the association of CTSS and RDX-FDX1 with atherosclerotic immune cells. Tissue/cell type enrichment analysis showed the involvement of arteries, adrenal glands, and fat tissues in the development of CAD. We next compared the odds ratios of lead variants for myocardial infarction at 76 genome-wide significant loci in the transethnic meta-analysis and a moderate correlation between Japanese and Europeans, where 8 loci showed a difference. Finally, we performed tissue/cell type enrichment analysis using East Asian-frequent and European-frequent variants according to the risk allele frequencies and identified significant enrichment of adrenal glands in the East Asian-frequent group while the enrichment of arteries and fat tissues was found in the European-frequent group. These findings indicate biological differences in CAD susceptibility between Japanese and Europeans.

CONCLUSIONS

We identified 3 new loci for CAD and highlighted the genetic differences between the Japanese and European populations. Moreover, our transethnic analyses showed both shared and unique genetic architectures between the Japanese and Europeans. While most of the underlying genetic bases for CAD are shared, further analyses in diverse populations will be needed to elucidate variations fully.

Complete Resolution of Left Ventricular Outflow Tract Obstruction After Spontaneous Mitral Valve Chordal Rupture in a Patient With Hypertrophic Cardiomyopathy

•

Spontaneous mitral chordal rupture is a complication in hypertrophic cardiomyopathy (HCM).

•

Mitral chordal rupture in HCM causes deterioration in heart failure.

•

Symptoms improved when left ventricular outflow tract (LVOT) obstruction disappeared.

•

Mitral valve has a role in LVOT obstruction and systolic anterior motion.

A new virus related to Satsuma dwarf virus: the nucleotide sequence of the 3'-terminal regions of Hyuganatsu virus RNAs 1 and 2. Brief Report

The sequences of the 3'-terminal 1306 and 2160 nucleotides of RNAs 1 and 2 of a virus serologically related to Satsuma dwarf virus (SDV) from Hyuganatsu ( Citrus tamurana Hort. ex Tan.) were determined, respectively. We found that the partial RNA-dependent RNA polymerase region in RNA1 and the coat proteins (CPs) region in RNA2 of the virus tentatively named Hyuganatsu virus (HV) have 78.3-84.0% and 76.9-80.7% amino acid sequence identities to those of known SDV-related viruses (SDV-RVs), i.e., SDV, Citrus mosaic virus, and Navel orange infectious mottling virus. Sequence analyses show that HV is classifiable as a new SDV-RV species.

Nucleotide sequence and taxonomy of Cycas necrotic stunt virus. Brief report

Cycas necrotic stunt virus (CNSV) is the only well-characterized virus from gymnosperm. cDNA segments corresponding to the bipartite genome RNAs (RNA1, RNA2) were synthesized and sequenced. Each RNA encoded a single polyprotein, flanked by the 5' and 3' non-coding regions (NCR) and followed by a poly (A) tail. The putative polyproteins encoded by RNA1 and RNA2 had sets of motifs, which were characteristic of viruses in the genus Nepovirus. The polyproteins showed higher sequence identities to Artichoke Italian latent virus, Grapevine chrome mosaic virus and Tomato black ring virus, all of which belong to subgroup b of the genus Nepovirus, than to other nepoviruses. Phylogenetic analysis of RNA dependent RNA polymerase and coat protein also showed closer relationships with these viruses than other viruses. The data obtained supported the taxonomical status of CNSV as a definitive member of the genus Nepovirus, subgroup b.

Characterization of a new citrus viroid species tentatively termed Citrus viroid OS

A new viroid was obtained from a viroid sample, named 'OS', collected from a citrus plant. The viroid consists of 330-331 nucleotides, contains the central conserved region (CCR) characteristic of the genus Apscaviroid, and has the highest sequence similarity (only 68%) with Citrus III viroid (CVd-III) among known viroids. This viroid, by itself, caused only mild petiole necrosis and characteristically very mild leaf bending in Arizona 861-S1 Etrog citrons (Citrus medica L.), the degree of which differed from that induced by other citrus viroids. This viroid could be a new citrus viroid species belonging to the genus Apscaviroid; for convenience, it was tentatively named Citrus viroid OS (CVd-OS) after the original sample. CVd-OS has chimeric features related to other viroids. In particular, CVd-OS has high sequence similarity with CVd-III and Apple dimple fruit viroid in the putative central and terminal left domains, including a duplicative sequence from the lower CCR of the genus Pospiviroid in the left terminus. Further, CVd-OS shares high sequence similarity with Citrus exocortis viroid (CEVd) in the lower strand of the putative variable and terminal right domains, including the sequence identical to that of the right termini of CEVd and of Citrus IV viroid.

A population of variants of a viroid closely related to citrus viroid-I in citrus plants

Sequencing analyses showed a population of variants consisting of 325-330 nucleotides (nt) of a viroid closely related to citrus viroid (CVd)-I in citrus plants. These variants, for which we propose the tentative acronym CVd-I-LSS (low sequence similarity), have only 82-85% sequence similarities to CVd-I variants. A phylogenetic tree showed that the CVd-I-LSS variants formed an individual cluster that was distinct from that of the CVd-I variants, but no intermediate variants were found which could continuously connect the population of CVd-I-LSS variants with that of CVd-I variants. Citrons (Citrus medica L.) inoculated with the CVd-I-LSS seemed to show not only moderate leaf bending like citrons infected with CVd-I but also slightly more severe epinasty than citrons with CVd-I. Other biological properties of CVd-I-LSS, such as the host range, will need to be determined in order to clarify whether CVd-I-LSS is a new viroid species or a distinct strain of CVd-I. Most of the nucleotide changes between the CVd-I-LSS and CVd-I variants were found at complementary positions of the upper and lower strands within the putative pathogenic, variable, and right terminal domains, as well as in the region surrounding the central conserved region of the predicted secondary structure of CVd-I-LSS.

Infection Behavior of Venturia nashicola, the Cause of Scab on Asian Pears

ABSTRACT The infection of Japanese pear by Venturia nashicola, the cause of scab on Asian pears (Japanese pear, Pyrus pylifolia var. culta; Chinese pear, P. ussuriensis), was examined using light and electron microscopy to determine the mechanism of resistance in pears. Early stages of infection were similar on the susceptible cv. Kosui, the resistant cv. Kinchaku, and the nonhost European pear (P. communis) cv. Flemish Beauty. V. nashicola penetrated only the cuticle layer on pear leaves and formed subcuticular hyphae on all three cultivars. Hyphae were localized in the pectin layer of pear leaves and never penetrated into the cytoplasm of epidermal cells. This restriction of fungal growth suggested that pectinases released by infection hyphae or subcuticular hyphae may be important in infection. Subcuticular hyphae were modified ultrastructurally in the pectin layer of resistant pear cultivars accompanied by fungal cell death. In contrast, fungal cells appeared intact in susceptible pear cultivars, suggesting the existence of resistance mechanisms.

Transgenic grapevine plants expressing a rice chitinase with enhanced resistance to fungal pathogens

 The rice chitinase gene (RCC2), classified as class I chitinase, was introduced into the somatic embryos of grapevine (Vitis vinifera L. cv. Neo Muscut) by Agrobacterium infection. After co-cultivation with Agrobacterium, somatic embryos were transferred onto Murashige and Skoog hormone-free medium supplemented with 50 mg/l kanamycin. Transformed secondary or tertiary embryos were selected, and then more than 20 transgenic plantlets were recovered. Two transformants showed enhanced resistance against powdery mildew caused by Uncinula necator. Few disease symptoms were observed on leaves of these transformants compared with those of the non-transformant, although browning and necrotic symptoms, which seemed to constitute a hypersensitive reaction, were observed. Scanning electron microscopic observation revealed that conidial germination, mycelial growth and conidial formation were suppressed on the leaf surface of the transformant. The transgenic grapevines obtained also exhibited slight resistance against Elisinoe ampelina inducing anthracnose, resulting in a reduction in disease lesions. The relationship between the expression of the foreign chitinase gene and the disease resistance is discussed.

Citrus viroid Ia is a derivative of citrus bent leaf viroid (CVd-Ib) by partial sequence duplications in the right terminal region

Nucleotide sequences of group I citrus viroids Ia (CVd-Ia) and citrus bent leaf viroid (CBLVd, formerly designated CVd-Ib) isolated from citrus plants in Japan, the Philippines and China have been determined. Citrus samples in Japan and the Philippines contained CVd-Ia, which consists of 328 nucleotides(nt). Although 10 nt longer than the type CBLVd-225A in Israel they share 94% identity in overall nucleotide sequence. The Philippines sample also contained a 329-nt long CVd-Ia sequence variant, in which one base insertion and three substitutions were observed. A citrus in China contained CBLVd, which consists of 318 nt and shares 98% identity to CBLVd-225A. CVd-Ia was clearly separated from CBLVd by two 5-nt insertions located in upper (5'-AGCUG-3') and the lower (5'-CUUCU-3') strand of the right terminal region (which is also designated T2 domain) in rod-like secondary structure. Since both of the additional 5-nt sequences are similar to the adjacent sequences (5'-AGUUG-3' and 5'-CUUCU-3'), we hypothesize that CVd-Ia is a derivative of CBLVd caused by partial sequence duplications and substitutions taking place in the right terminal region.

The nucleotide sequence of the coat protein genes of satsuma dwarf virus and naval orange infectious mottling virus

The sequence of the 3'-terminal 4320 and 2409 nucleotides were determined for RNA2 of satsuma dwarf virus (SDV) and navel infectious mottling virus (NIMV). Both sequences contained a part of a long open reading frame which encodes larger and smaller coat proteins (CPs) at the 3'-terminus followed by a 3'non-coding region upstream of a poly (A) tail. Amino acid sequence identity for larger and smaller CPs ranged 81-84% and 68-78%, respectively, among SDV, NIMV and the previously sequenced citrus mosaic virus (CiMV). No significant sequence similarity was found between the CPs of SDV or NIMV and those of the como-, nepo- or other viruses. The nucleotide sequence identity of the 3' non-coding region of RNA2 were 68%-78% among SDV, CiMV and NIMV. These results suggest that SDV, CiMV and NIMV are distinct, though related, viruses. They may be assigned as members of the new genus, which is close to the genera of Comovirus and Nepovirus.

Nucleotide sequence of the coat protein genes of citrus mosaic virus

The sequence of the 3'-terminal 2,201 nucleotides of RNA2 of citrus mosaic virus (CiMV) was determined. The sequence contains a long open reading frame (ORF) of 1989 nucleotides in the virus-sense strand and at 3' untranslated region of 212 nucleotides upstream of a poly(A) tail. The N-terminal amino acid sequence of the two coat proteins was determined by Edman degradation and the corresponding coding regions were identified in the polyprotein. The larger coat protein with Mr 48,122 is encoded upstream of the smaller one with Mr 24,172. The coat proteins are apparently cleaved from the polyprotein at an Arg-Gly and a Thr-Asn bond. Although CiMV has properties in common with the comoviruses and nepoviruses, there is no significant sequence homology between the coat proteins of CiMV and those of either group. Furthermore, the coat proteins of CiMV lack homology with those of strawberry latent ringspot virus (SLRSV), which has been reported to be more distantly related to the como- and nepoviruses. This lack of homology distinguishes CiMV from the como- and nepoviruses, SLRSV and other viruses.

PCR detection of Ti and Ri plasmids from phytopathogenic Agrobacterium strains

A universal primer set (VCF/VCR) for PCR analysis based on the sequences of the virC operon located on Ti and Ri plasmids was designed to detect these plasmids from phytopathogenic Agrobacterium strains. With the VCF (sequence, 5'-ATCATTTGTAGCGACT-3') and VCR (sequence, 5'-AGCTCAAACCTGCTTC-3') primer set, DNA fragments of 730 bp in length were amplified from cell lysates of 10 rhizogenic and 65 tumorigenic agrobacteria. DNA sequencing and Southern hybridization analysis confirmed that the amplified fragments corresponded to the target region. The PCR method is considered convenient for routine determination of the potential pathogenicity of Agrobacterium strains.

Proposal for rejection of Agrobacterium tumefaciens and revised descriptions for the genus Agrobacterium and for Agrobacterium radiobacter and Agrobacterium rhizogenes

The 16S rRNA sequences of seven representative Agrobacterium strains, eight representative Rhizobium strains, and the type strains of Azorhizobium caulinodans and Bradyrhizobium japonicum were determined. These strains included the type strains of Agrobacterium tumefaciens, Agrobacterium rhizogenes, Agrobacterium radiobacter, Agrobacterium vitis, Agrobacterium rubi, Rhizobium fredii, Rhizobium galegae, Rhizobium huakuii, Rhizobium leguminosarum, Rhizobium loti, Rhizobium meliloti, and Rhizobium tropici. A phylogenetic analysis showed that the 15 strains of Agrobacterium and Rhizobium species formed a compact phylogenetic cluster clearly separated from the other members of the alpha subclass of the Proteobacteria. However, Agrobacterium species and Rhizobium species are phylogenetically entwined with one another, and the two genera cannot be separated. In the Agrobacterium species, the strains of biovar 1, biovar 2, Agrobacterium rubi, and Agrobacterium vitis were clearly separated. The two biovars exhibited homogeneity in their phenotypic, chemotaxonomic, and phylogenetic characteristics, and two species should be established for the two biovars. We considered the nomenclature of the two biovars, and revised descriptions of Agrobacterium radiobacter (for the biovar 1 strains) and Agrobacterium rhizogenes (for the biovar 2 strains) are proposed. The name Agrobacterium tumefaciens is rejected because the type strain of this species was assigned to Agrobacterium radiobacter, and consequently the description of the genus Agrobacterium is revised.