Improving Generalizability of PET DL Algorithms: List-Mode Reconstructions Improve DOTATATE PET Hepatic Lesion Detection Performance

Deep learning (DL) algorithms used for DOTATATE PET lesion detection typically require large, well-annotated training datasets. These are difficult to obtain due to low incidence of gastroenteropancreatic neuroendocrine tumors (GEP-NETs) and the high cost of manual annotation. Furthermore, networks trained and tested with data acquired from site specific PET/CT instrumentation, acquisition and processing protocols have reduced performance when tested with offsite data. This lack of generalizability requires even larger, more diverse training datasets. The objective of this study is to investigate the feasibility of improving DL algorithm performance by better matching the background noise in training datasets to higher noise, out-of-domain testing datasets. 68Ga-DOTATATE PET/CT datasets were obtained from two scanners: Scanner1, a state-of-the-art digital PET/CT (GE DMI PET/CT; n = 83 subjects), and Scanner2, an older-generation analog PET/CT (GE STE; n = 123 subjects). Set1, the data set from Scanner1, was reconstructed with standard clinical parameters (5 min; Q.Clear) and list-mode reconstructions (VPFXS 2, 3, 4, and 5-min). Set2, data from Scanner2 representing out-of-domain clinical scans, used standard iterative reconstruction (5 min; OSEM). A deep neural network was trained with each dataset: Network1 for Scanner1 and Network2 for Scanner2. DL performance (Network1) was tested with out-of-domain test data (Set2). To evaluate the effect of training sample size, we tested DL model performance using a fraction (25%, 50% and 75%) of Set1 for training. Scanner1, list-mode 2-min reconstructed data demonstrated the most similar noise level compared that of Set2, resulting in the best performance (F1 = 0.713). This was not significantly different compared to the highest performance, upper-bound limit using in-domain training for Network2 (F1 = 0.755; p-value = 0.103). Regarding sample size, the F1 score significantly increased from 25% training data (F1 = 0.478) to 100% training data (F1 = 0.713; p < 0.001). List-mode data from modern PET scanners can be reconstructed to better match the noise properties of older scanners. Using existing data and their associated annotations dramatically reduces the cost and effort in generating these datasets and significantly improves the performance of existing DL algorithms. List-mode reconstructions can provide an efficient, low-cost method to improve DL algorithm generalizability.

Learning Without Real Data Annotations to Detect Hepatic Lesions in PET Images

OBJECTIVE

Deep neural networks have been recently applied to lesion identification in fluorodeoxyglucose (FDG) positron emission tomography (PET) images, but they typically rely on a large amount of well-annotated data for model training. This is extremely difficult to achieve for neuroendocrine tumors (NETs), because of low incidence of NETs and expensive lesion annotation in PET images. The objective of this study is to design a novel, adaptable deep learning method, which uses no real lesion annotations but instead low-cost, list mode-simulated data, for hepatic lesion detection in real-world clinical NET PET images.

METHODS

We first propose a region-guided generative adversarial network (RG-GAN) for lesion-preserved image-to-image translation. Then, we design a specific data augmentation module for our list-mode simulated data and incorporate this module into the RG-GAN to improve model training. Finally, we combine the RG-GAN, the data augmentation module and a lesion detection neural network into a unified framework for joint-task learning to adaptatively identify lesions in real-world PET data.

RESULTS

The proposed method outperforms recent state-of-the-art lesion detection methods in real clinical 68Ga-DOTATATE PET images, and produces very competitive performance with the target model that is trained with real lesion annotations.

CONCLUSION

With RG-GAN modeling and specific data augmentation, we can obtain good lesion detection performance without using any real data annotations.

SIGNIFICANCE

This study introduces an adaptable deep learning method for hepatic lesion identification in NETs, which can significantly reduce human effort for data annotation and improve model generalizability for lesion detection with PET imaging.

Location-Aware Encoding for Lesion Detection in 68Ga-DOTATATE Positron Emission Tomography Images

OBJECTIVE

Lesion detection with positron emission tomography (PET) imaging is critical for tumor staging, treatment planning, and advancing novel therapies to improve patient outcomes, especially for neuroendocrine tumors (NETs). Current lesion detection methods often require manual cropping of regions/volumes of interest (ROIs/VOIs) a priori, or rely on multi-stage, cascaded models, or use multi-modality imaging to detect lesions in PET images. This leads to significant inefficiency, high variability and/or potential accumulative errors in lesion quantification. To tackle this issue, we propose a novel single-stage lesion detection method using only PET images.

METHODS

We design and incorporate a new, plug-and-play codebook learning module into a U-Net-like neural network and promote lesion location-specific feature learning at multiple scales. We explicitly regularize the codebook learning with direct supervision at the network's multi-level hidden layers and enforce the network to learn multi-scale discriminative features with respect to predicting lesion positions. The network automatically combines the predictions from the codebook learning module and other layers via a learnable fusion layer.

RESULTS

We evaluate the proposed method on a real-world clinical 68Ga-DOTATATE PET image dataset, and our method produces significantly better lesion detection performance than recent state-of-the-art approaches.

CONCLUSION

We present a novel deep learning method for single-stage lesion detection in PET imaging data, with no ROI/VOI cropping in advance, no multi-stage modeling and no multi-modality data.

SIGNIFICANCE

This study provides a new perspective for effective and efficient lesion identification in PET, potentially accelerating novel therapeutic regimen development for NETs and ultimately improving patient outcomes including survival.

Learning with limited target data to detect cells in cross-modality images

Deep neural networks have achieved excellent cell or nucleus quantification performance in microscopy images, but they often suffer from performance degradation when applied to cross-modality imaging data. Unsupervised domain adaptation (UDA) based on generative adversarial networks (GANs) has recently improved the performance of cross-modality medical image quantification. However, current GAN-based UDA methods typically require abundant target data for model training, which is often very expensive or even impossible to obtain for real applications. In this paper, we study a more realistic yet challenging UDA situation, where (unlabeled) target training data is limited and previous work seldom delves into cell identification. We first enhance a dual GAN with task-specific modeling, which provides additional supervision signals to assist with generator learning. We explore both single-directional and bidirectional task-augmented GANs for domain adaptation. Then, we further improve the GAN by introducing a differentiable, stochastic data augmentation module to explicitly reduce discriminator overfitting. We examine source-, target-, and dual-domain data augmentation for GAN enhancement, as well as joint task and data augmentation in a unified GAN-based UDA framework. We evaluate the framework for cell detection on multiple public and in-house microscopy image datasets, which are acquired with different imaging modalities, staining protocols and/or tissue preparations. The experiments demonstrate that our method significantly boosts performance when compared with the reference baseline, and it is superior to or on par with fully supervised models that are trained with real target annotations. In addition, our method outperforms recent state-of-the-art UDA approaches by a large margin on different datasets.

Genomic Signature for Oligometastatic Disease in Non-Small Cell Lung Cancer Patients with Brain Metastases

PURPOSE/OBJECTIVE(S)

Biomarkers for oligometastatic disease remain elusive and few studies have attempted to correlate genomic data to the presence of true oligometastatic disease.

MATERIALS/METHODS

Patients with non-small cell lung cancer (NSCLC) and brain metastases were identified in our departmental database. Electronic medical records were used to identify patients for whom liquid biopsy-based comprehensive genomic profiling (Guardant Health) was available. Oligometastatic disease was defined as patients having ≤5 non-brain metastases without diffuse involvement of a single organ. Widespread disease was any spread beyond oligometastatic. Fisher's exact tests were used to identify mutations statistically associated (p<0.1) with either oligometastatic or widespread extracranial disease. A score of +1 was assigned for every mutation present associated with oligometastatic disease, and -1 was assigned for mutations associated with widespread disease. Scores were summed for each patient to create a risk score for the likelihood of oligometastatic disease, with scores subsequently correlated to the likelihood of having oligometastatic disease vs widespread disease. For oligometastatic patients, a competing risk analysis was done to assess for cumulative incidence of oligometastatic progression accounting for the potential competing risks of widespread progression of extracranial disease or death. Cox regression was used to determine the association between oligometastatic risk score and oligometastatic progression.

RESULTS

One hundred thirty patients met study criteria and were included in the analysis. 51 patients (39%) had oligometastatic disease. Genetic mutations included in the Guardant panel associated (p<0.1) with the presence of oligometastatic extracranial disease included ATM, JAK2, MAP2K2, and NTRK1; ARID1A and CCNE1 were associated with widespread disease. Patients with a positive, neutral and negative risk score for oligometastatic disease had a 78%, 41% and 11.5% likelihood of having oligometastatic disease, respectively (p<0.0001). Overall survival for patients with positive, neutral and negative risk scores for oligometastatic disease was 86% vs 82% vs 64% at 6 months (p = 0.2). The competing risk analysis found that the oligometastatic risk score was significantly associated with the likelihood of oligometastatic progression based on the Wald Chi-square test. Patients with positive, neutral and negative risk scores for oligometastatic disease had a cumulative incidence of oligometastatic progression of 77% vs 35% vs 33% at 6 months (p = 0.03 from competing risk model).

CONCLUSION

Elucidation of a genomic signature for oligometastatic disease derived from non-invasive liquid biopsy appears feasible for NSCLC patients. Patients with the oligometastatic signature exhibited higher rates of early oligometastatic progression. Validation of this signature could lead to a biomarker that has the potential to direct local therapies in oligometastatic patients.

Joint Cranial Bone Labeling and Landmark Detection in Pediatric CT Images Using Context Encoding

Image segmentation, labeling, and landmark detection are essential tasks for pediatric craniofacial evaluation. Although deep neural networks have been recently adopted to segment cranial bones and locate cranial landmarks from computed tomography (CT) or magnetic resonance (MR) images, they may be hard to train and provide suboptimal results in some applications. First, they seldom leverage global contextual information that can improve object detection performance. Second, most methods rely on multi-stage algorithm designs that are inefficient and prone to error accumulation. Third, existing methods often target simple segmentation tasks and have shown low reliability in more challenging scenarios such as multiple cranial bone labeling in highly variable pediatric datasets. In this paper, we present a novel end-to-end neural network architecture based on DenseNet that incorporates context regularization to jointly label cranial bone plates and detect cranial base landmarks from CT images. Specifically, we designed a context-encoding module that encodes global context information as landmark displacement vector maps and uses it to guide feature learning for both bone labeling and landmark identification. We evaluated our model on a highly diverse pediatric CT image dataset of 274 normative subjects and 239 patients with craniosynostosis (age 0.63 ± 0.54 years, range 0-2 years). Our experiments demonstrate improved performance compared to state-of-the-art approaches.

[Comparative analysis of intestinal microbiota distribution characteristics based on metagenomics in patients with hepatitis B cirrhosis with or without ascites]

Objective: To use metagenomic sequencing to compare the differences in intestinal microbiota species and metabolic pathways in patients with hepatitis B cirrhosis with or without ascites and further explore the correlation between the differential microbiota and clinical indicators and metabolic pathways. Methods: 20 hepatitis B cirrhosis cases [10 without ascites (HBLC-WOA), 10 with ascites (HBLC-WA), and 5 healthy controls (HC)] were selected from the previously studied 16S rRNA samples. Metagenome sequencing was performed on the intestinal microbiota samples. The Kruskal-Wallis rank sum test and Spearman test were used to identify and analyse differential intestinal microbiota populations, metabolic pathways, and their correlations. Results: (1) The overall structure of the intestinal microbiota differed significantly among the three groups (R = 0.19, P = 0.018). The HC group had the largest abundance of Firmicutes and the lowest abundance of Proteobacteria at the genus level. Firmicutes abundance was significantly decreased (P(fdr) < 0.01), while Proteobacteria abundance was significantly increased (P(fdr) < 0.01) in patients with cirrhosis accompanied by ascites; (2) LEfSe analysis revealed that 29 intestinal microbiota (18 in the HBLC-WA group and 11 in the HBLC-WOA group) played a significant role in the disease group. The unclassified Enterobacteriaceae and Klebsiella species in the HBLC-WA group and Enterobacteriaceae in the HBLC-WOA group were positively correlated with the Child-Turcotte-Pugh (CTP) score, prothrombin time, and international normalized ratio score and negatively correlated with albumin and hemoglobin levels (P < 0.05). Escherichia and Shigella in the HBLC-WA group were positively correlated with CTP scores (P < 0.05); (3) The correlation analysis results between the KEGG pathway and 29 specific intestinal microbiota revealed that Enterobacteriaceae and arachidonic acid, α-linolenic acid, glycerolipid metabolism, and fatty acid degradation were positively correlated in the lipid metabolism pathway, while most Enterobacteriaceae were positively correlated with branched-chain amino acid degradation and negatively correlated with aromatic amino acid biosynthesis in the amino acid metabolic pathway. Conclusion: A significant increment of Enterobacteriaceae in the intestines of HBLC-WA patients influenced hepatic reserve function and was associated with amino acid and lipid metabolic pathways. Therefore, attention should be paid to controlling the intestinal microbiota to prevent complications and improve the prognosis in patients with hepatitis B cirrhosis, especially in those with ascites.

A platform-independent framework for phenotyping of multiplex tissue imaging data

Multiplex imaging is a powerful tool to analyze the structural and functional states of cells in their morphological and pathological contexts. However, hypothesis testing with multiplex imaging data is a challenging task due to the extent and complexity of the information obtained. Various computational pipelines have been developed and validated to extract knowledge from specific imaging platforms. A common problem with customized pipelines is their reduced applicability across different imaging platforms: Every multiplex imaging technique exhibits platform-specific characteristics in terms of signal-to-noise ratio and acquisition artifacts that need to be accounted for to yield reliable and reproducible results. We propose a pixel classifier-based image preprocessing step that aims to minimize platform-dependency for all multiplex image analysis pipelines. Signal detection and noise reduction as well as artifact removal can be posed as a pixel classification problem in which all pixels in multiplex images can be assigned to two general classes of either I) signal of interest or II) artifacts and noise. The resulting feature representation maps contain pixel-scale representations of the input data, but exhibit significantly increased signal-to-noise ratios with normalized pixel values as output data. We demonstrate the validity of our proposed image preprocessing approach by comparing the results of two well-accepted and widely-used image analysis pipelines.

Deep learning on graphs for multi-omics classification of COPD

Network approaches have successfully been used to help reveal complex mechanisms of diseases including Chronic Obstructive Pulmonary Disease (COPD). However despite recent advances, we remain limited in our ability to incorporate protein-protein interaction (PPI) network information with omics data for disease prediction. New deep learning methods including convolution Graph Neural Network (ConvGNN) has shown great potential for disease classification using transcriptomics data and known PPI networks from existing databases. In this study, we first reconstructed the COPD-associated PPI network through the AhGlasso (Augmented High-Dimensional Graphical Lasso Method) algorithm based on one independent transcriptomics dataset including COPD cases and controls. Then we extended the existing ConvGNN methods to successfully integrate COPD-associated PPI, proteomics, and transcriptomics data and developed a prediction model for COPD classification. This approach improves accuracy over several conventional classification methods and neural networks that do not incorporate network information. We also demonstrated that the updated COPD-associated network developed using AhGlasso further improves prediction accuracy. Although deep neural networks often achieve superior statistical power in classification compared to other methods, it can be very difficult to explain how the model, especially graph neural network(s), makes decisions on the given features and identifies the features that contribute the most to prediction generally and individually. To better explain how the spectral-based Graph Neural Network model(s) works, we applied one unified explainable machine learning method, SHapley Additive exPlanations (SHAP), and identified CXCL11, IL-2, CD48, KIR3DL2, TLR2, BMP10 and several other relevant COPD genes in subnetworks of the ConvGNN model for COPD prediction. Finally, Gene Ontology (GO) enrichment analysis identified glycosaminoglycan, heparin signaling, and carbohydrate derivative signaling pathways significantly enriched in the top important gene/proteins for COPD classifications.

[The effect of lncRNA ADPGK-AS1 on the proliferation and apoptosis of retinoblastoma cells by targeting miR-200b-5p]

Objective: To explore the effect of lncRNA ADPGK-AS1 on the proliferation and apoptosis of retinoblastoma cells and its possible mechanism. Methods: The tumor tissues of 31 patients with retinoblastoma admitted to Henan Provincial Eye Hospital from February to June 2020 and their corresponding normal tissues adjacent to the cancer were collected. The expression levels of lncRNA ADPGK-AS1 and miR-200b-5p in retinoblastoma tissues and normal adjacent tissues were detected by real-time fluorescence quantitative polymerase chain reaction (qRT-PCR). Human retinal epithelial cell ARPE-19, human retinoblastoma cell Y-79 and WERI-Rb-1 were cultured in vitro. The expression levels of lncRNA ADPGK-AS1 and miR-200b-5p were detected by qRT-PCR. Y-79 cells were randomly divided into si-con group, si-lncRNA ADPGK-AS1 group, miR con group, miR-200b-5p group, si-lncRNA ADPGK-AS1+ anti-miR con group, and si-lncRNA ADPGK-AS1+ anti-miR-200b-5p group. The proliferation, cloning and apoptosis of cells in each group were detected by tetramethylazol blue method, plate cloning test and flow cytometry, respectively. The targeting relationship between lncRNA ADPGK-AS1 and miR-200b-5p was detected by double luciferase report test, and the expression level of cleaved-caspase-3 protein was detected by western blot. Results: Compared with the adjacent tissues, the expression of lncRNA ADPGK-AS1 in retinoblastoma tissues was increased (P<0.05), while the expression of miR-200b-5p was decreased (P<0.05). Compared with ARPE-19 cells, the expression of lncRNA ADPGK-AS1 in Y-79 and WERI-Rb-1 cells was increased (P<0.05), while the expression of miR-200b-5p was decreased (P<0.05). Compared with the si-con group, the cell viability of the si-lncRNA ADPGK-AS1 group was reduced (1.06±0.09 vs 0.53±0.05, P<0.05), the number of cell clone formation was reduced (114.00±8.03 vs 57.00±4.13, P<0.05), while the apoptosis rate [(7.93±0.68)% vs (25.43±1.94)%] and the protein level of cleaved-caspase-3 were increased (P<0.05). Compared with the miR-con group, the cell viability of the miR-200b-5p group was decreased (1.05±0.08 vs 0.57±0.05, P<0.05), the number of cell clone formation was decreased (118.00±10.02 vs 64.00±5.13, P<0.05), while the apoptosis rate [(7.89±0.71)% vs (23.15±1.62)%] and the protein level of cleaved-caspase-3 were increased (P<0.05). lncRNA ADPGK-AS1 could target the expression of miR-200b-5p. Compared with the si-lncRNA ADPGK-AS1+ anti-miR-con group, cell viability of the si-lncRNA ADPGK-AS1+ anti-miR-200b-5p group was increased (0.53±0.04 vs 1.25±0.10, P<0.05), and the number of cell clones was increased (54.00±4.39 vs 125.00±10.03, P<0.05), while the rate of apoptosis [(25.38±1.53)% vs (9.76±0.71)%] and the protein level of cleaved-caspase-3 were decreased (P<0.05). Conclusion: Interfering with the expression of lncRNA ADPGK-AS1 could inhibit the proliferation and clone formation and induce apoptosis of retinoblastoma cells by targeting the expression of miR-200b-5p.

Modeling contrast-to-noise ratio from list mode reconstructions of 68Ga DOTATATE PET/CT: predicting detectability of hepatic metastases in shorter acquisition PET reconstructions

BACKGROUND

Deep learning (DL) algorithms have shown promise in identifying and quantifying lesions in PET/CT. However, the accuracy and generalizability of these algorithms relies on large, diverse datasets which are time and labor intensive to curate. Modern PET/CT scanners may acquire data in list mode, allowing for multiple reconstructions of the same datasets with different parameters and imaging times. These reconstructions may provide a wide range of image characteristics to increase the size and diversity of datasets. Training algorithms with shorter imaging times and higher noise properties requires that lesions remain detectable. The purpose of this study is to model and predict the contrast-to-noise ratio (CNR) for shorter imaging times based on CNR from longer duration, lower noise images for ⁶⁸Ga DOTATATE PET hepatic lesions and identify a threshold above which lesions remain detectable.

METHODS

⁶⁸Ga DOTATATE subjects (n=20) with hepatic lesions were divided into two subgroups. The "Model" group (n=4 subjects; n=9 lesions; n=36 datapoints) was used to identify the relationship between CNR and imaging time. The "Test" group (n=16 subjects; n=44 lesions; n=176 datapoints) was used to evaluate the prediction provided by the model.

RESULTS

CNR plotted as a function of imaging time for a subset of identified subjects was very well fit with a quadratic model. For the remaining subjects, the measured CNR showed a very high linear correlation with the predicted CNR for these lesions (R² > 0.97) for all imaging durations. From the model, a threshold CNR=6.9 at 5-minutes predicted CNR > 5 at 2-minutes. Visual inspection of lesions in 2-minute images with CNR above the threshold in 5-minute images were assessed and rated as a 4 or 5 (probably positive or definitely positive) confirming 100% lesion detectability on the shorter 2-minute PET images.

CONCLUSIONS

CNR for shorter DOTATATE PET imaging times may be accurately predicted using list mode reconstructions of longer acquisitions. A threshold CNR may be applied to longer duration images to ensure lesion detectability of shorter duration reconstructions. This method can aid in the selection of lesions to include in novel data augmentation techniques for deep learning.

Low-Resource Adversarial Domain Adaptation for Cross-Modality Nucleus Detection

Due to domain shifts, deep cell/nucleus detection models trained on one microscopy image dataset might not be applicable to other datasets acquired with different imaging modalities. Unsupervised domain adaptation (UDA) based on generative adversarial networks (GANs) has recently been exploited to close domain gaps and has achieved excellent nucleus detection performance. However, current GAN-based UDA model training often requires a large amount of unannotated target data, which may be prohibitively expensive to obtain in real practice. Additionally, these methods have significant performance degradation when using limited target training data. In this paper, we study a more realistic yet challenging UDA scenario, where (unannotated) target training data is very scarce, a low-resource case rarely explored for nucleus detection in previous work. Specifically, we augment a dual GAN network by leveraging a task-specific model to supplement the target-domain discriminator and facilitate generator learning with limited data. The task model is constrained by cross-domain prediction consistency to encourage semantic content preservation for image-to-image translation. Next, we incorporate a stochastic, differentiable data augmentation module into the task-augmented GAN network to further improve model training by alleviating discriminator overfitting. This data augmentation module is a plug-and-play component, requiring no modification of network architectures or loss functions. We evaluate the proposed low-resource UDA method for nucleus detection on multiple public cross-modality microscopy image datasets. With a single training image in the target domain, our method significantly outperforms recent state-of-the-art UDA approaches and delivers very competitive or superior performance over fully supervised models trained with real labeled target data.

[Effects of subanesthetic dose of esketamine on opioid consumption after thoracoscopic surgery]

Objective: To investigate the effect of continuous intravenous infusion of subanesthetic dose of esketamine intraoperatively on postoperative opioid consumption in patients undergoing thoracoscopic surgery. Methods: A total of 71 patients with elective thoracoscopic lung surgery in the First Affiliated Hospital of Zhengzhou University from December 2020 to December 2021 were selected. Patients who were classified as grade Ⅰ or Ⅱ by the American Society of Anesthesiologists (ASA) and aged 18-70 years were included, including 32 males and 39 females, with a body mass index (BMI) of 18.5-30.0 kg/m². The patients were randomly divided into three groups: (1) Control group (group C, n=24): continuous intravenous infusion of normal saline at the same rate during surgery; (2) Subanesthetic dose of esketamine 0.125 mg·kg^-1·h^-1 group (group ES1, n=23): continuous intravenous infusion of esketamine at a rate of 0.125 mg·kg^-1·h^-1 during surgery; (3) Subanesthetic dose of esketamine 0.250 mg·kg^-1·h^-1 group (group ES2, n=24): continuous intravenous infusion of esketamine at a rate of 0.250 mg·kg^-1·h^-1 during surgery. The main outcome measures were the total consumptions of hydromorphone of 3 groups within 24 and 48 hours after surgery. The secondary outcome measures were the extubation time, length of postanesthesia care unit (PACU) stay, the time of first feeding, and the incidences of adverse effects within 24 h after surgery in 3 groups. Results: The 24 h postoperative consumption of hydromorphone in group C, ES1 and ES2 was (5.4±1.0) mg, (4.5±1.5) mg and (4.0±0.8) mg, respectively. Likewise, the 48 h postoperative consumption of hydromorphone was (9.7±2.2) mg, (9.0±3.0) mg and (7.7±1.8) mg, respectively. Compared with group C, the 24 h postoperative hydromorphone consumptions were significantly reduced in group ES1 and ES2 (both P<0.05). The extubation time, length of PACU stay and the time of first feeding after surgery in group C were (23±10) min,(70±12) min,(17±3) h,in group ES1 were (22±4) min,(69±11) min,(14±5) h,in group ES2 were (16±8) min,(58±12) min,(14±3) h, respectively. Compared with group C and group ES1, both of the extubation time and length of PACU stay were shortened in group ES2 (both P<0.05). Compared with group C, the first postoperative feeding time of group ES1 and ES2 was shortened (both P<0.05). There were no differences in the incidences of adverse effects at postoperative 24 h among 3 groups (all P>0.05). Conclusion: Continuously intravenous infusion of subanesthetic esketamine at a rate of 0.250 mg·kg^-1·h^-1 can significantly reduce the postoperative opioid consumption and improve the patient's outcomes.

[Characteristics of fusion gene expression in acute lymphoblastic leukemia]

Objective: To analyze the genetic landscape of 52 fusion genes in patients with de novo acute lymphoblastic leukemia (ALL) and to investigate the characteristics of other laboratory results. Methods: The fusion gene expression was retrospectively analyzed in the 1 994 patients with de novo ALL diagnosed from September 2016 to December 2020. In addition, their mutational, immunophenotypical and karyotypical profiles were investigated. Results: In the 1 994 patients with ALL, the median age was 12 years (from 15 days to 89 years). In the panel of targeted genes, 15 different types of fusion genes were detected in 884 patients (44.33%) and demonstrated a Power law distribution. The frequency of detectable fusion genes in B-cell ALL was significantly higher than that in T-cell ALL (48.48% vs 18.71%), and fusion genes were almost exclusively expressed in B-cell ALL or T-cell ALL. The number of fusion genes showed peaks at<1 year, 3-5 years and 35-44 years, respectively. More fusion genes were identified in children than in adults. MLL-FG was most frequently seen in infants and TEL-AML1 was most commonly seen in children, while BCR-ABL1 was dominant in adults. The majority of fusion gene mutations involved signaling pathway and the most frequent mutations were observed in NRAS and KRAS genes. The expression of early-stage B-cell antigens varied in B-cell ALL patients. The complex karyotypes were more common in BCR-ABL1 positive patients than others. Conclusion: The distribution of fusion genes in ALL patients differs by ages and cell lineages. It also corresponds to various gene mutations, immunophenotypes, and karyotypes.

Cluster activation mapping with application to computed tomography scans of the lung

Purpose: An open question in deep clustering is how to explain what in the image is driving the cluster assignments. This is especially important for applications in medical imaging when the derived cluster assignments may inform decision-making or create new disease subtypes. We develop cluster activation mapping (CLAM), which is methodology to create localization maps highlighting the image regions important for cluster assignment. Approach: Our approach uses a linear combination of the activation channels from the last layer of the encoder within a pretrained autoencoder. The activation channels are weighted by a channelwise confidence measure, which is a modification of score-CAM. Results: Our approach performs well under medical imaging-based simulation experiments, when the image clusters differ based on size, location, and intensity of abnormalities. Under simulation, the cluster assignments were predicted with 100% accuracy when the number of clusters was set at the true value. In addition, applied to computed tomography scans from a sarcoidosis population, CLAM identified two subtypes of sarcoidosis based purely on CT scan presentation, which were significantly associated with pulmonary function tests and visual assessment scores, such as ground-glass, fibrosis, and honeycombing. Conclusions: CLAM is a transparent methodology for identifying explainable groupings of medical imaging data. As deep learning networks are often criticized and not trusted due to their lack of interpretability, our contribution of CLAM to deep clustering architectures is critical to our understanding of cluster assignments, which can ultimately lead to new subtypes of diseases.

An Augmented High-Dimensional Graphical Lasso Method to Incorporate Prior Biological Knowledge for Global Network Learning

Biological networks are often inferred through Gaussian graphical models (GGMs) using gene or protein expression data only. GGMs identify conditional dependence by estimating a precision matrix between genes or proteins. However, conventional GGM approaches often ignore prior knowledge about protein-protein interactions (PPI). Recently, several groups have extended GGM to weighted graphical Lasso (wGlasso) and network-based gene set analysis (Netgsa) and have demonstrated the advantages of incorporating PPI information. However, these methods are either computationally intractable for large-scale data, or disregard weights in the PPI networks. To address these shortcomings, we extended the Netgsa approach and developed an augmented high-dimensional graphical Lasso (AhGlasso) method to incorporate edge weights in known PPI with omics data for global network learning. This new method outperforms weighted graphical Lasso-based algorithms with respect to computational time in simulated large-scale data settings while achieving better or comparable prediction accuracy of node connections. The total runtime of AhGlasso is approximately five times faster than weighted Glasso methods when the graph size ranges from 1,000 to 3,000 with a fixed sample size (n = 300). The runtime difference between AhGlasso and weighted Glasso increases when the graph size increases. Using proteomic data from a study on chronic obstructive pulmonary disease, we demonstrate that AhGlasso improves protein network inference compared to the Netgsa approach by incorporating PPI information.

Inhibition of Aspergillus flavus growth and aflatoxin B1 production by natamycin

Aspergillus flavus causes huge crop losses, reduces crop quality and has adverse effects on human and animal health. A large amount of food contaminated with aflatoxin can greatly increase the risk of liver cancer. Therefore, prevention and control of aflatoxin production have aroused attention of research in various countries. Natamycin extracted from Streptomyces spp. has been widely used in production practice due to its good specificity and safety. Here, we found that natamycin could significantly inhibit fungal growth, conidia germination, ergosterol and AFB1 production by A. flavus in a dose-dependent manner. Scanning electron microscope analysis indicated that the number of conidia was decreased, the outer wall of conidia was destroyed, and the mycelia were shrivelled and tangled by natamycin. RNA-Seq data indicated that natamycin inhibited fungal growth and conidia development of A. flavus by significantly down-regulating some genes involved in ergosterol biosynthesis, such as Erg13, HMG1 and HMG2. It inhibited conidia germination by significantly down-regulating some genes related to conidia development, such as FluG and VosA. After natamycin exposure, the decreased ratio of aflS/aflR caused by the down-regulation of all the structural genes, which subsequently resulted in the suppression of AFB1 production. In conclusion, this study served to reveal the inhibitory mechanisms of natamycin on fungal growth and AFB1 biosynthesis in A. flavus and to provide solid evidence for its application in controlling AFB1 contamination.

Automated liver lesion detection in 68Ga DOTATATE PET/CT using a deep fully convolutional neural network

BACKGROUND

Gastroenteropancreatic neuroendocrine tumors most commonly metastasize to the liver; however, high normal background ⁶⁸Ga-DOTATATE activity and high image noise make metastatic lesions difficult to detect. The purpose of this study is to develop a rapid, automated and highly specific method to identify ⁶⁸Ga-DOTATATE PET/CT hepatic lesions using a 2D U-Net convolutional neural network.

METHODS

A retrospective study of ⁶⁸Ga-DOTATATE PET/CT patient studies (n = 125; 57 with ⁶⁸Ga-DOTATATE hepatic lesions and 68 without) was evaluated. The dataset was randomly divided into 75 studies for the training set (36 abnormal, 39 normal), 25 for the validation set (11 abnormal, 14 normal) and 25 for the testing set (11 abnormal, 14 normal). Hepatic lesions were physician annotated using a modified PERCIST threshold, and boundary definition by gradient edge detection. The 2D U-Net was trained independently five times for 100,000 iterations using a linear combination of binary cross-entropy and dice losses with a stochastic gradient descent algorithm. Performance metrics included: positive predictive value (PPV), sensitivity, F₁ score and area under the precision-recall curve (PR-AUC). Five different pixel area thresholds were used to filter noisy predictions.

RESULTS

A total of 233 lesions were annotated with each abnormal study containing a mean of 4 ± 2.75 lesions. A pixel filter of 20 produced the highest mean PPV 0.94 ± 0.01. A pixel filter of 5 produced the highest mean sensitivity 0.74 ± 0.02. The highest mean F₁ score 0.79 ± 0.01 was produced with a 20 pixel filter. The highest mean PR-AUC 0.73 ± 0.03 was produced with a 15 pixel filter.

CONCLUSION

Deep neural networks can automatically detect hepatic lesions in ⁶⁸Ga-DOTATATE PET. Ongoing improvements in data annotation methods, increasing sample sizes and training methods are anticipated to further improve detection performance.

Bidirectional Mapping-Based Domain Adaptation for Nucleus Detection in Cross-Modality Microscopy Images

Cell or nucleus detection is a fundamental task in microscopy image analysis and has recently achieved state-of-the-art performance by using deep neural networks. However, training supervised deep models such as convolutional neural networks (CNNs) usually requires sufficient annotated image data, which is prohibitively expensive or unavailable in some applications. Additionally, when applying a CNN to new datasets, it is common to annotate individual cells/nuclei in those target datasets for model re-learning, leading to inefficient and low-throughput image analysis. To tackle these problems, we present a bidirectional, adversarial domain adaptation method for nucleus detection on cross-modality microscopy image data. Specifically, the method learns a deep regression model for individual nucleus detection with both source-to-target and target-to-source image translation. In addition, we explicitly extend this unsupervised domain adaptation method to a semi-supervised learning situation and further boost the nucleus detection performance. We evaluate the proposed method on three cross-modality microscopy image datasets, which cover a wide variety of microscopy imaging protocols or modalities, and obtain a significant improvement in nucleus detection compared to reference baseline approaches. In addition, our semi-supervised method is very competitive with recent fully supervised learning models trained with all real target training labels.

Generative Adversarial Domain Adaptation for Nucleus Quantification in Images of Tissue Immunohistochemically Stained for Ki-67

PURPOSE

We focus on the problem of scarcity of annotated training data for nucleus recognition in Ki-67 immunohistochemistry (IHC)–stained pancreatic neuroendocrine tumor (NET) images. We hypothesize that deep learning–based domain adaptation is helpful for nucleus recognition when image annotations are unavailable in target data sets.

METHODS

We considered 2 different institutional pancreatic NET data sets: one (ie, source) containing 38 cases with 114 annotated images and the other (ie, target) containing 72 cases with 20 annotated images. The gold standards were manually annotated by 1 pathologist. We developed a novel deep learning–based domain adaptation framework to count different types of nuclei (ie, immunopositive tumor, immunonegative tumor, nontumor nuclei). We compared the proposed method with several recent fully supervised deep learning models, such as fully convolutional network-8s (FCN-8s), U-Net, fully convolutional regression network (FCRN) A, FCRNB, and fully residual convolutional network (FRCN). We also evaluated the proposed method by learning with a mixture of converted source images and real target annotations.

RESULTS

Our method achieved an F₁ score of 81.3% and 62.3% for nucleus detection and classification in the target data set, respectively. Our method outperformed FCN-8s (53.6% and 43.6% for nucleus detection and classification, respectively), U-Net (61.1% and 47.6%), FCRNA (63.4% and 55.8%), and FCRNB (68.2% and 60.6%) in terms of F₁ score and was competitive with FRCN (81.7% and 70.7%). In addition, learning with a mixture of converted source images and only a small set of real target labels could further boost the performance.

CONCLUSION

This study demonstrates that deep learning–based domain adaptation is helpful for nucleus recognition in Ki-67 IHC stained images when target data annotations are not available. It would improve the applicability of deep learning models designed for downstream supervised learning tasks on different data sets.

CircRNA_100395 protects breast carcinoma deterioration by targeting MAPK6

OBJECTIVE

This study aims to uncover the differential expression of circRNA_100395 in breast carcinoma specimens, and its regulatory effect on cancer cell phenotypes. The role of circRNA_100395 in affecting breast carcinoma progression and the molecular mechanism are explored as well.

PATIENTS AND METHODS

CircRNA_100395 expressions in breast carcinoma and paracancerous tissues were detected. The influence of circRNA_100395 level on clinical indicators of breast carcinoma patients was analyzed. In vitro regulations of circRNA_100395 on phenotypes of breast carcinoma cells were examined by CCK-8, colony formation, and transwell assay. The interaction between circRNA_100395 and MAPK6 was confirmed by Dual-Luciferase reporter assay and rescue assays.

RESULTS

CircRNA_100395 was downregulated in breast carcinoma tissues and cell lines. Its level was negatively correlated to tumor staging and tumor size of breast carcinoma. Overexpression of circRNA_100395 in SKBR3 and MDA-MB-231 cells weakened proliferative and migratory abilities. MAPK6 was the target gene of circRNA_100395. Overexpression of MAPK6 reversed the anti-cancer effect of circRNA_100395 on breast carcinoma.

CONCLUSIONS

CircRNA_100395 serves as an anti-cancer gene in breast carcinoma progression by targeting MAPK6, and its level is negatively correlated to tumor staging and tumor size of breast carcinoma. CircRNA_100395 can be utilized as a potential biomarker and therapeutic target of breast carcinoma.

[Bimatoprost promotes hair growth of reconstructed hair follicles in mice through activation of the Wnt/β-catenin signaling pathway]

Objective: To investigate effect of Bimatoprost (BimP) on growth of reconstructed hair follicles in recipient nude mice. Methods: Primary epidermal and dermal cells were isolated from newborn C57BL/6J mice (1-day-old) skins, and the reconstructed hair follicles was implanted in the dorsal skin of Balb/c-nu nude mice using a silicon chamber protocol, then, the 18 nude mice were randomly divided into control group, BimP group and minoxidil group, with 6 mice in each group. After 2 weeks, topical treatment was applied to the grafted area of the nude mice by 2% minoxidil 100 μl, 0.03% BimP 100 μl and saline 100 μl, respectively, once daily for 2 weeks. At day 14 after treatment, the mice were euthanized to measure the length of dorsal hair, and the number and hair cycle of the reconstructed follicles was observed histologically. The total mRNA and proteins expression of Wnt3a, LEF1, β-catenin and Frizzled7 were determined by qPCR and Western Blotting. The distribution and expression of β-catenin in the reconstructed follicles was detected by immunofluorescence staining. Results: As compared to the control group, the BimP group had thicker and longer hair [(0.57±0.07) vs (0.36±0.05) cm, P<0.01], no significant difference was seen between the BimP and minoxidil group. The mRNA expression levels of Wnt3a (2.73±0.17 vs 1.00±0.14, P<0.01)、LEF1(1.71±0.12 vs 1.00±0.19, P<0.01)、β-catenin (2.37±0.21vs 1.00±0.11, P<0.01) and Frizzled7 (2.62±0.15vs 1.00±0.18, P<0.01) were significantly increased in BimP group compared with the control group. Western Blotting showed the same results, the protein expression levels of Wnt3a (1.44±0.21vs 1.00±0.13, P<0.05)、LEF1 (1.36±0.15 vs 1.00±0.09, P<0.05)、β-catenin (1.60±0.13 vs 1.00±0.16, P<0.01) and Frizzled7 (1.52±0.15 vs 1.00±0.21, P<0.05) in BimP group were higher than those in control group, and the difference was statistically significant. Immunofluorescence staining showed that β-catenin was strongly expressed in hair bulb cells and sebaceous gland cells of reconstructed hair follicles in BimP group and minoxidil group, whereas barely seen in the control group. Conclusion: BimP directly promotes growth of reconstructed hair follicles in mice by activating canonical Wnt/β-catenin signaling pathway.

A rapid sap-direct reverse transcription-polymerase chain reaction method for detection of dendrobium viroid in Dendrobium plants

Dendrobium viroid (DVd) was first reported in China in 2020, and it is the only viroid known to infect Orchidaceae family plants. In this study, we developed a simple reverse transcription-polymerase chain reaction (RT-PCR) method for the rapid detection of DVd in Dendrobium plants. When extracting the sap template from the leaves, they are first clamped between two layers of plastic film, and the sap is pressed out and collected with a pipette. Using this sap, DVd was detected by dot-blot and RT-PCR methods and, the expected amplicons were confirmed by sequencing analysis. The batch analysis of field samples revealed that this method can be used to detect DVd rapidly. The detection method also reduces cross-contamination between different samples and minimizes false positives. Thus, this sap-direct RT-PCR method allows effective and rapid DVd detection in the study of Orchidaceae plants.

Loss-Based Attention for Interpreting Image-Level Prediction of Convolutional Neural Networks

Although deep neural networks have achieved great success on numerous large-scale tasks, poor interpretability is still a notorious obstacle for practical applications. In this paper, we propose a novel and general attention mechanism, loss-based attention, upon which we modify deep neural networks to mine significant image patches for explaining which parts determine the image decision-making. This is inspired by the fact that some patches contain significant objects or their parts for image-level decision. Unlike previous attention mechanisms that adopt different layers and parameters to learn weights and image prediction, the proposed loss-based attention mechanism mines significant patches by utilizing the same parameters to learn patch weights and logits (class vectors), and image prediction simultaneously, so as to connect the attention mechanism with the loss function for boosting the patch precision and recall. Additionally, different from previous popular networks that utilize max-pooling or stride operations in convolutional layers without considering the spatial relationship of features, the modified deep architectures first remove them to preserve the spatial relationship of image patches and greatly reduce their dependencies, and then add two convolutional or capsule layers to extract their features. With the learned patch weights, the image-level decision of the modified deep architectures is the weighted sum on patches. Extensive experiments on large-scale benchmark databases demonstrate that the proposed architectures can obtain better or competitive performance to state-of-the-art baseline networks with better interpretability. The source codes are available on: https://github.com/xsshi2015/Loss-based-Attention-for-Interpreting-Image-level-Prediction-of-Convolutional-Neural-Networks.

A Scalable Optimization Mechanism for Pairwise Based Discrete Hashing

Maintaining the pairwise relationship among originally high-dimensional data into a low-dimensional binary space is a popular strategy to learn binary codes. One simple and intuitive method is to utilize two identical code matrices produced by hash functions to approximate a pairwise real label matrix. However, the resulting quartic problem in term of hash functions is difficult to directly solve due to the non-convex and non-smooth nature of the objective. In this paper, unlike previous optimization methods using various relaxation strategies, we aim to directly solve the original quartic problem using a novel alternative optimization mechanism to linearize the quartic problem by introducing a linear regression model. Additionally, we find that gradually learning each batch of binary codes in a sequential mode, i.e. batch by batch, is greatly beneficial to the convergence of binary code learning. Based on this significant discovery and the proposed strategy, we introduce a scalable symmetric discrete hashing algorithm that gradually and smoothly updates each batch of binary codes. To further improve the smoothness, we also propose a greedy symmetric discrete hashing algorithm to update each bit of batch binary codes. Moreover, we extend the proposed optimization mechanism to solve the non-convex optimization problems for binary code learning in many other pairwise based hashing algorithms. Extensive experiments on benchmark single-label and multi-label databases demonstrate the superior performance of the proposed mechanism over recent state-of-the-art methods on two kinds of retrieval tasks: similarity and ranking order. The source codes are available on https://github.com/xsshi2015/Scalable-Pairwise-based-Discrete-Hashing.

Depression and one-year survival of patients with heart failure in China: analysis from the China-PEACE Prospective Heart Failure study

Background

Depression is prevalent among patients with heart failure (HF), but data exploring association of depression with risk of death in patients with HF is scarce in China. We investigated the relationship between depression and all-cause mortality of heart failure in China.

Methods

In China PEACE 5p-HF Study, we prospectively enrolled patients primarily hospitalized with HF from 52 diverse hospitals throughout China during 2016–2018. All the patients were followed up for 1 year. About 10% patients in the cohort from 41 hospitals was included for the measurement of depression state at convenience. Depression was measured by the Patient Health Questionnaire-8 depression scale (PHQ-8) at baseline. Depression state was categorized into major depressive disorder (10–24 points), minor depression (5–10 points) and no depression (0–5 points). Cox proportional hazards regression analyses, controlling for established risk factors as age, gender, LVEF, NYHA, medication use and medical history, were used to evaluate how depression were related to end point of death from any cause.

Results

Total 584 patients were included in our analysis, with median age 69 (IQR 60–77) years, and 40.8% female. Among these patients, 36.0% had major depressive disorder (n=210), 33.9% had minor depression (n=198). There were 70 (12%) patients died within 1 year after discharge. Major depressive disorder was associated with higher all-cause mortality compared with no depression (hazard ratio=2.18, 95% confidence interval 1.36–3.50, p=0.001). While minor depression was not significantly associated with all-cause mortality.

Conclusions

Major depression is an independent risk factor for all-cause mortality in hospitalized patients with HF in China. It is necessary to screen for psychological health in hospitalized patients to targeting intervention.

Funding Acknowledgement

Type of funding source: Public Institution(s). Main funding source(s): National Key Research and Development Program from the Ministry of Science and Technology of China

Discharge heart rate and clinical outcomes in hospitalized heart failure patients with coexisted atrial fibrillation

Background

Whether discharge heart rate for hospitalized heart failure (HF) patients with coexisted atrial fibrillation (AF) is associated with long-term clinical outcomes and whether this association differs between patients with and without beta-blockers have not been well studied.

Purpose

We investigated the associations between discharge heart rate and clinical outcomes in hospitalized HF patients with coexisted AF, while stratified to beta-blockers at discharge.

Methods

The study cohort included 1631 HF patients hospitalized primarily with AF, which was from the China PEACE Prospective Heart Failure Study. Clinical outcome was 1-year combined all-cause mortality and HF hospitalization after discharge. We analyzed association between outcome and heart rate at discharge with restricted cubic spline and Cox proportional hazard ratios (HR).

Results

The median age was 68 (IQR: 60- 77) years, 41.9% were women, discharge heart rate was (median (IQR)) 75 (69- 84) beats per minute (bpm), and 60.2% received beta-blockers at discharge. According to the result of restricted cubic spline plot, the relationship between discharge heart rate and clinical outcome may be nonlinear (P&lt;0.01). Based on above result, these patients were divided into 3 groups: lowest &lt;65 bpm, middle 65–86 bpm and highest ≥87 bpm, clinical outcomes occurred in 128 (64.32%), 624 (53.42%) and 156 (59.32%) patients in the lowest, middle, and highest groups respectively. In the Cox proportional hazard analysis, the lowest and highest groups were associated with increased risks of clinical outcome compared with the middle group (HR: 1.289, 95% confidence interval (CI): 1.056 - 1.573, p=0.013; HR: 1.276, 95% CI: 1.06 - 1.537, p=0.01, respectively). And a significant interaction between discharge heart rate and beta-blocker use was observed (P&lt;0.001 for interaction). Stratified analysis showed the lowest group was associated with increased risks of clinical outcomes in patients with beta-blockers (HR: 1.584, 95% confidence interval (CI): 1.215–2.066, p=0.001).

Conclusion

There may be a U-curve relationship between discharge heart rate and clinical outcomes in hospitalized HF patients with coexisted AF. They may have the best clinical outcomes with heart rates of 65 - 86 bpm. And strict heart rate control (&lt;65 bpm) may be avoided for patients who discharge with beta-blockers.

Figure 1

Funding Acknowledgement

Type of funding source: Other. Main funding source(s): This work was supported by the National Key Research and Development Program (2017YFC1310803) from the Ministry of Science and Technology of China; the CAMS Innovation Fund for Medical Science (2017-I2M-B&R-02); the 111 Project from the Ministry of Education of China (B16005).

Impacts of beta-blockers on lone term clinical outcomes in the treatment of hospitalized heart failure patients with an ejection fraction greater than 40%

Background

Heart failure (HF) is a leading cause of mortality and morbidity. Beta-blocker is recommended in HF with reduced ejection fraction (EF) in order to improve clinical outcomes. While the effects of beta-blockers in HF who have an EF greater than 40% are uncertain and controversial yet.

Purpose

To Investigate the associations between beta-blockers and clinical outcomes, overall and in strata of patients with an EF of between 40% and 49% or greater than 50%.

Methods

The study cohort included 2642 HF patients hospitalized primarily for HF who had an EF greater than 40%, which was from the China PEACE Prospective Heart Failure Study. We had two Clinical outcomes: 1-year all-cause mortality and 1-year hospitalization for HF after discharge. The associations between beta-blockers and clinical outcomes were assessed using Cox proportional hazard regression models, while stratified according to EF.

Results

The median age was 70 (IQR: 61, 77) years, 44.8% were women, EF was (median (IQR)) 54% (46%, 62%), and 55.5% received beta-blockers at discharge. All-cause mortality and hospitalizations for HF occurred in 341 (12.91%) and 636 (24.07%) patients respectively. In the Cox proportional hazard analysis, a significant interaction between EF and beta-blocker use for mortality was observed (P=0.01 for interaction). Stratified analysis showed beta-blockers reduced risks of mortality in patients who had an EF between 40% and 49% ((hazard ratios (HR): 0.501, 95% confidence interval (CI): 0.340- 0.738, p&lt;0.001), but not among patients with an EF of 50% or greater (HR: 0.824, 95% CI: 0.600- 1.133, p=0.233). Use of β-blockers was not associated with reduced hospitalizations in patients with EF of between 40% and 49% and greater than 50% (HR: 1.016, 95% CI: 0.712- 1.450, p=0.931; HR: 0.905, 95% CI: 0.703- 1.166, p=0.439, respectively).

Conclusion

For patients with an EF between 40% and 49%, β-blocker use was associated with a reduced risk of all-cause mortality but not HF hospitalizations. For patients with an EF of 50% or greater, there was no such association.

Funding Acknowledgement

Type of funding source: Other. Main funding source(s): This work was supported by the National Key Research and Development Program (2017YFC1310803) from the Ministry of Science and Technology of China; the CAMS Innovation Fund for Medical Science (2017-I2M-B&R-02); the 111 Project from the Ministry of Education of China (B16005).

Discharge heart rate and beta-blockers treatments in hospitalized heart failure patients with coexisted atrial fibrillation

Background

Beta-blockers are widely used to improve clinical outcomes in heart failure (HF) patients. However, the effects of beta-blockers on clinical outcomes in those who have coexisted atrial fibrillation (AF) remains uncertain.

Purpose

We investigated the associations between beta-blockers and clinical outcomes according to discharge heart rate.

Methods

The study cohort included 1631 HF patients hospitalized primarily with AF, which was from the China PEACE Prospective Heart Failure Study. Clinical outcome was 1-year combined all-cause mortality and HF hospitalization after discharge. The associations between beta-blockers and clinical outcome were assessed using Cox proportional hazard and standardization mortality weighting regression models, with stratified discharge heart rate group predefined by restricted cubic spline.

Results

The median age was 68 (IQR: 60- 77) years, 41.9% were women, discharge heart rate was (median (IQR)) 75 (69- 84) beats per minute (bpm), and 60.2% received beta-blockers at discharge. According to the result of restricted cubic spline plot, these patients were divided into 3 groups: lowest &lt;65 bpm, middle 65–86 bpm and highest ≥87 bpm (Fig.1). In the Cox proportional hazard analysis, a significant interaction between discharge heart rate and beta-blocker use was observed (P&lt;0.001 for interaction). Stratified analysis showed beta-blocker prescription at discharge was associated with reduced risk for clinical outcomes in patients with high heart rates (hazard ratio 0.336, 95% CI: 0.144–0.786, p=0.012) but not in those with lowest and middle heart rates (hazard ratio: 1.32; 95% CI, 0.95–1.63; hazard ratio: 1.02; 95% CI, 0.68–1.55, respectively).

Conclusion

The associations between beta-blockers and clinical outcomes may be significantly influenced by baseline heart rate. Hospitalized HF patients with AF benefit the most from beta-blockers use if they had high heart rate (≥87 bpm) at discharge.

Figure 1

Funding Acknowledgement

Type of funding source: Other. Main funding source(s): This work was supported by the National Key Research and Development Program (2017YFC1310803) from the Ministry of Science and Technology of China; the CAMS Innovation Fund for Medical Science (2017-I2M-B&R-02); the 111 Project from the Ministry of Education of China (B16005).

Quantifying the incremental value of deep learning: Application to lung nodule detection

We present a case study for implementing a machine learning algorithm with an incremental value framework in the domain of lung cancer research. Machine learning methods have often been shown to be competitive with prediction models in some domains; however, implementation of these methods is in early development. Often these methods are only directly compared to existing methods; here we present a framework for assessing the value of a machine learning model by assessing the incremental value. We developed a machine learning model to identify and classify lung nodules and assessed the incremental value added to existing risk prediction models. Multiple external datasets were used for validation. We found that our image model, trained on a dataset from The Cancer Imaging Archive (TCIA), improves upon existing models that are restricted to patient characteristics, but it was inconclusive about whether it improves on models that consider nodule features. Another interesting finding is the variable performance on different datasets, suggesting population generalization with machine learning models may be more challenging than is often considered.

[Observation of the therapeutic and characteristic effects of terlipressin on refractory cirrhotic ascites]

Objective: To observe the therapeutic effect of terlipressin on refractory ascites (RA) in cirrhosis, and its role and impact on acute kidney injury (AKI). Methods: A non-randomized controlled clinical trial data of 111 hospitalized cases of liver cirrhosis accompanied with RA was collected from Shuguang Hospital Affiliated to Shanghai University of Traditional Chinese Medicine, Zhongshan Hospital of Hubei Province, The First Affiliated Hospital of Zhengzhou University, The First Affiliated Hospital of Medical School of Zhejiang University, and People's Hospital of Pudong New Area of Shanghai between March 2015 and March 2017. 26 cases of conventional treatment group (control group) were divided into two subgroups: RA without AKI (RA-NAKI) and RA with AKI (RA-AKI), and each subgroup consisted 13 cases. Patients with bacterial infection were treated with diuretics, albumin supplementation and antibiotics. 85 cases were presented in terlipressin combined treatment group, of which 27 cases were of RA-NAKI and 58 cases were of RA-AKI. Control group was injected terlipressin 1mg of intravenous drip or static push (once q6 h ~ 12 h) for more than 5 days. The treatment duration lasted for 2 weeks with 4 weeks of follow-up. Body weight, 24-hour urine volume, abdominal circumference, mean arterial pressure (MAP), liver and kidney function, anterior hepatic ascites, deepest point of ascites, and ultrasonographic detection of ascites in supine position before treatment, one and two weeks after treatment and 4 weeks after follow-up were compared. Count data were tested by χ (2). Samples of four groups at baseline were compared. One-way analysis of variance was used for normal distribution data and Kruskal-Wallis H test for non-normal distribution data. Repeated measures analysis of variance was used to compare the difference in efficacy between different time points before and after treatment in the group. The LSD method of one-way ANOVA was used to compare the two groups. A t-test of independent samples was used to compare the efficacy of different time series between the two groups. Mann-Whitney rank- sum test was used to compare the data of non-normal distribution between the two groups. Results: (1) Baseline data were compared among 4 subgroups of terlipressin RA-NAKI and control RA-AKI. Control group age was higher than that of terlipressin group, and the serum creatinine (SCr) of the RA-AKI group was higher than RA-NAKI group, and there was no significant difference in the rest of the baseline data and the combined medication (P > 0.05). (2) An intra-group comparison between control and trelipressin before and after treatment showed that all patients had lower body mass, abdominal circumference and deepest ascites, and higher serum albumin (P < 0.05). 24-hour urine volume and MAP was significantly increased in the terlipressin group, while the pre-ascites, SCr and child Turcotte Pugh (CTP) scores were decreased. Body weight, abdominal circumference, pre-ascites, and deepest ascites of the terlipressin group were decreased, while MAP was increased during the treatment and follow-up periods. The differences were statistically significant when compared with the control group at the same time (P < 0.05). There was a statistically significant difference in the increase of 24-h urine volume in the terlipressin group compared with the control group (P < 0.05). The decrease in SCr and CTP in the terlipressin group after 2 weeks of treatment and 4 weeks of follow-up was statistically significant compared with the control group (P < 0.05). (3) Among the two subgroups of RA-AKI and RA-NAKI in the terlipressin group, the baseline SCr value of the former was higher than that of the latter. After treatment, the body weight, abdominal circumference, pre-ascites, deepest ascites and CTP scores were decreased. In the RA-AKI group, 24-hour urine volume, MAP, SCr and serum albumin concentration were significantly increased. The difference between the two subgroups before and after treatment was compared, and the body weight of RA-AKI group at 1, 2 weeks of treatment and 4 weeks of follow-up was significantly lower than RA-NAKI group, which were (- 2.3 ± 0.2 vs. - 1.5 ± 0.2) kg, (- 4.1 ± 0.2 vs. - 2.6 ± 0.2) kg, (- 4.2 ± 0.3 vs. - 2.4 ± 0.3) kg, respectively. RA-NAKI group urine volume was significantly increased at 2 weeks of treatment and 4 weeks of follow-up, which was (468 ± 42 vs. 110 ± 131) ml, (272 ± 34 ml vs. 11 ± 112) ml, respectively. SCr reduction (18.3 ± 4.7 vs. 0.9 ± 2.4) µmol/l at 4 weeks of follow-up was apparent in RA-NAKI group, and the difference was statistically significant (P < 0.05). Conclusion: Addition of terlipressin to conventional treatment may significantly increase MAP, 24-h urine volume, improve renal function and promote ascites resolution in patients with refractory cirrhotic ascites. Moreover, its combination effect is more obvious in AKI patients, and adverse reactions are mild.

Assessment of a Machine Learning Model Applied to Harmonized Electronic Health Record Data for the Prediction of Incident Atrial Fibrillation

IMPORTANCE

Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, and its early detection could lead to significant improvements in outcomes through the appropriate prescription of anticoagulation medication. Although a variety of methods exist for screening for AF, a targeted approach, which requires an efficient method for identifying patients at risk, would be preferred.

OBJECTIVE

To examine machine learning approaches applied to electronic health record data that have been harmonized to the Observational Medical Outcomes Partnership Common Data Model for identifying risk of AF.

DESIGN, SETTING, AND PARTICIPANTS

This diagnostic study used data from 2 252 219 individuals cared for in the UCHealth hospital system, which comprises 3 large hospitals in Colorado, from January 1, 2011, to October 1, 2018. Initial analysis was performed in December 2018; follow-up analysis was performed in July 2019.

EXPOSURES

All Observational Medical Outcomes Partnership Common Data Model-harmonized electronic health record features, including diagnoses, procedures, medications, age, and sex.

MAIN OUTCOMES AND MEASURES

Classification of incident AF in designated 6-month intervals, adjudicated retrospectively, based on area under the receiver operating characteristic curve and F1 statistic.

RESULTS

Of 2 252 219 individuals (1 225 533 [54.4%] women; mean [SD] age, 42.9 [22.3] years), 28 036 (1.2%) developed incident AF during a designated 6-month interval. The machine learning model that used the 200 most common electronic health record features, including age and sex, and random oversampling with a single-layer, fully connected neural network provided the optimal prediction of 6-month incident AF, with an area under the receiver operating characteristic curve of 0.800 and an F1 score of 0.110. This model performed only slightly better than a more basic logistic regression model composed of known clinical risk factors for AF, which had an area under the receiver operating characteristic curve of 0.794 and an F1 score of 0.079.

CONCLUSIONS AND RELEVANCE

Machine learning approaches to electronic health record data offer a promising method for improving risk prediction for incident AF, but more work is needed to show improvement beyond standard risk factors.

Graph temporal ensembling based semi-supervised convolutional neural network with noisy labels for histopathology image analysis

Although convolutional neural networks have achieved tremendous success on histopathology image classification, they usually require large-scale clean annotated data and are sensitive to noisy labels. Unfortunately, labeling large-scale images is laborious, expensive and lowly reliable for pathologists. To address these problems, in this paper, we propose a novel self-ensembling based deep architecture to leverage the semantic information of annotated images and explore the information hidden in unlabeled data, and meanwhile being robust to noisy labels. Specifically, the proposed architecture first creates ensemble targets for feature and label predictions of training samples, by using exponential moving average (EMA) to aggregate feature and label predictions within multiple previous training epochs. Then, the ensemble targets within the same class are mapped into a cluster so that they are further enhanced. Next, a consistency cost is utilized to form consensus predictions under different configurations. Finally, we validate the proposed method with extensive experiments on lung and breast cancer datasets that contain thousands of images. It can achieve 90.5% and 89.5% image classification accuracy using only 20% labeled patients on the two datasets, respectively. This performance is comparable to that of the baseline method with all labeled patients. Experiments also demonstrate its robustness to small percentage of noisy labels.

Adversarial Domain Adaptation and Pseudo-Labeling for Cross-Modality Microscopy Image Quantification

Cell or nucleus quantification has recently achieved state-of-the-art performance by using convolutional neural networks (CNNs). In general, training CNNs requires a large amount of annotated microscopy image data, which is prohibitively expensive or even impossible to obtain in some applications. Additionally, when applying a deep supervised model to new datasets, it is common to annotate individual cells in those target datasets for model re-training or fine-tuning, leading to low-throughput image analysis. In this paperSSS, we propose a novel adversarial domain adaptation method for cell/nucleus quantification across multimodality microscopy image data. Specifically, we learn a fully convolutional network detector with task-specific cycle-consistent adversarial learning, which conducts pixel-level adaptation between source and target domains and completes a cell/nucleus detection task. Then we generate pseudo-labels on target training data using the detector trained with adapted source images and further fine-tune the detector towards the target domain to boost the performance. We evaluate the proposed method on multiple cross-modality microscopy image datasets and obtain a significant improvement in cell/nucleus detection compared to the reference baselines and a recent state-of-the-art deep domain adaptation approach. In addition, our method is very competitive with the fully supervised models trained with all real target training labels.

Towards pixel-to-pixel deep nucleus detection in microscopy images

BACKGROUND

Nucleus is a fundamental task in microscopy image analysis and supports many other quantitative studies such as object counting, segmentation, tracking, etc. Deep neural networks are emerging as a powerful tool for biomedical image computing; in particular, convolutional neural networks have been widely applied to nucleus/cell detection in microscopy images. However, almost all models are tailored for specific datasets and their applicability to other microscopy image data remains unknown. Some existing studies casually learn and evaluate deep neural networks on multiple microscopy datasets, but there are still several critical, open questions to be addressed.

RESULTS

We analyze the applicability of deep models specifically for nucleus detection across a wide variety of microscopy image data. More specifically, we present a fully convolutional network-based regression model and extensively evaluate it on large-scale digital pathology and microscopy image datasets, which consist of 23 organs (or cancer diseases) and come from multiple institutions. We demonstrate that for a specific target dataset, training with images from the same types of organs might be usually necessary for nucleus detection. Although the images can be visually similar due to the same staining technique and imaging protocol, deep models learned with images from different organs might not deliver desirable results and would require model fine-tuning to be on a par with those trained with target data. We also observe that training with a mixture of target and other/non-target data does not always mean a higher accuracy of nucleus detection, and it might require proper data manipulation during model training to achieve good performance.

CONCLUSIONS

We conduct a systematic case study on deep models for nucleus detection in a wide variety of microscopy images, aiming to address several important but previously understudied questions. We present and extensively evaluate an end-to-end, pixel-to-pixel fully convolutional regression network and report a few significant findings, some of which might have not been reported in previous studies. The model performance analysis and observations would be helpful to nucleus detection in microscopy images.

Deep Convolutional Hashing for Low-Dimensional Binary Embedding of Histopathological Images

Compact binary representations of histopa-thology images using hashing methods provide efficient approximate nearest neighbor search for direct visual query in large-scale databases. They can be utilized to measure the probability of the abnormality of the query image based on the retrieved similar cases, thereby providing support for medical diagnosis. They also allow for efficient managing of large-scale image databases because of a low storage requirement. However, the effectiveness of binary representations heavily relies on the visual descriptors that represent the semantic information in the histopathological images. Traditional approaches with hand-crafted visual descriptors might fail due to significant variations in image appearance. Recently, deep learning architectures provide promising solutions to address this problem using effective semantic representations. In this paper, we propose a deep convolutional hashing method that can be trained "point-wise" to simultaneously learn both semantic and binary representations of histopathological images. Specifically, we propose a convolutional neural network that introduces a latent binary encoding (LBE) layer for low-dimensional feature embedding to learn binary codes. We design a joint optimization objective function that encourages the network to learn discriminative representations from the label information, and reduce the gap between the real-valued low-dimensional embedded features and desired binary values. The binary encoding for new images can be obtained by forward propagating through the network and quantizing the output of the LBE layer. Experimental results on a large-scale histopathological image dataset demonstrate the effectiveness of the proposed method.

Pixel-to-Pixel Learning With Weak Supervision for Single-Stage Nucleus Recognition in Ki67 Images

OBJECTIVE

Nucleus recognition is a critical yet challenging step in histopathology image analysis, for example, in Ki67 immunohistochemistry stained images. Although many automated methods have been proposed, most use a multi-stage processing pipeline to categorize nuclei, leading to cumbersome, low-throughput, and error-prone assessments. To address this issue, we propose a novel deep fully convolutional network for single-stage nucleus recognition.

METHODS

Instead of conducting direct pixel-wise classification, we formulate nucleus identification as a deep structured regression model. For each input image, it produces multiple proximity maps, each of which corresponds to one nucleus category and exhibits strong responses in central regions of the nuclei. In addition, by taking into consideration the nucleus distribution in histopathology images, we further introduce an auxiliary task, region of interest (ROI) extraction, to assist and boost the nucleus quantification with weak ROI annotation. The proposed network can be learned in an end-to-end, pixel-to-pixel manner for simultaneous nucleus detection and classification.

RESULTS

We have evaluated this network on a pancreatic neuroendocrine tumor Ki67 image dataset, and the experiments demonstrate that our method outperforms recent state-of-the-art approaches.

CONCLUSION

We present a new, pixel-to-pixel deep neural network with two sibling branches for effective nucleus recognition and observe that learning with another relevant task, ROI extraction, can further boost individual nucleus localization and classification.

SIGNIFICANCE

Our method provides a clean, single-stage nucleus recognition pipeline for histopathology image analysis, especially a new perspective for Ki67 image quantification, which would potentially benefit individual object quantification in whole-slide images.

Texture Analysis for Muscular Dystrophy Classification in MRI with Improved Class Activation Mapping

The muscular dystrophies are made up of a diverse group of rare genetic diseases characterized by progressive loss of muscle strength and muscle damage. Since there is no cure for muscular dystrophy and clinical outcome measures are limited, it is critical to assess the progression of MD objectively. Imaging muscle replacement by fibrofatty tissue has been shown to be a robust biomarker to monitor disease progression in DMD. In magnetic resonance imaging (MRI) data, specific texture patterns are found to correlate to certain MD subtypes and thus present a potential way for automatic assessment. In this paper, we first apply state-of-the-art convolutional neural networks (CNNs) to perform accurate MD image classification and then propose an effective visualization method to highlight the important image textures. With a dystrophic MRI dataset, we found that the best CNN model delivers an 91.7% classification accuracy, which significantly outperforms non-deep learning methods, e.g., >40% improvement has been found over the traditional mean fat fraction (MFF) criterion for DMD and CMD classification. After investigating every single neuron at the top layer of CNN model, we found the superior classification ability of CNN can be explained by its 91 and 118 neurons were performing better than the MFF criterion under the measurements of Euclidean and Chi-square distance, respectively. In order to further interpret CNNs predictions, we tested an improved class activation mapping (ICAM) method to visualize the important regions in the MRI images. With this ICAM, CNNs are able to locate the most discriminative texture patterns of DMD in soleus, lateral gastrocnemius, and medial gastrocnemius; for CMD, the critical texture patterns are highlighted in soleus, tibialis posterior, and peroneus.

Treatment Response Prediction of Nasopharyngeal Carcinoma Based on Histogram Analysis of Diffusional Kurtosis Imaging

BACKGROUND AND PURPOSE

The prediction of treatment response is important in planning and modifying the chemoradiation therapy regimen. This study aimed to explore the quantitative histogram indices for treatment-response prediction of nasopharyngeal carcinoma based on diffusional kurtosis imaging compared with a standard ADC value (ADC_standard).

MATERIALS AND METHODS

Thirty-six patients with an initial diagnosis of locoregionally advanced nasopharyngeal carcinoma and diffusional kurtosis imaging acquisitions before and after neoadjuvant chemotherapy were enrolled. Patients were divided into respond-versus-nonrespond groups after neoadjuvant chemotherapy and residual-versus-nonresidual groups after radiation therapy. Histogram parameters of diffusional kurtosis imaging-derived parameters (ADC, ADC coefficient corrected by the non-Gaussain model [D], apparent kurtosis coefficient without a unit [K]) were calculated. The ADC_standard was calculated on the basis of intravoxel incoherent movement data. The intraclass correlation coefficient, Kolmogorov-Smirnov test, Student t test or Mann-Whitney U test, and receiver operating characteristic analysis were performed.

RESULTS

Most of the parameters had good-to-excellent consistency (intraclass correlation coefficient = 0.675-0.998). The pre-ADC_standard, pre-ADC (10th, 25th, 50th percentiles), pre-D (10th, 25th, 50th percentiles), and pre-K_50th were significantly different between the respond and nonrespond groups, while the pre-ADC_10th, pre-K_90th, post-ADC_50th, post-K_75th, post-K_90th, and the percentage change of parameters before and after neoadjuvant chemotherapy (▵ADC_50th%) were significantly different between the residual and nonresidual groups (all P < .05). Receiver operating characteristic analysis indicated that setting pre-D_50th = 0.875 × 10^-3mm²/s as the cutoff value could result in optimal diagnostic performance for neoadjuvant chemotherapy response prediction (area under the curve = 0.814, sensitivity = 0.70, specificity = 0.92), while the post-K_90th = 1.035 (area under the curve = 0.829, sensitivity = 0.78, specificity = 0.72), and▵ADC_50th% = 0.253 (area under the curve = 0.833, sensitivity = 0.94, specificity = 0.72) were optimal for radiation therapy response prediction.

CONCLUSIONS

Histogram analysis of diffusional kurtosis imaging may potentially predict the neoadjuvant chemotherapy and short-term radiation therapy response in locoregionally advanced nasopharyngeal carcinoma, therefore providing evidence for modification of the treatment regimen.