ResearchPad - deep-learning https://www.researchpad.co Default RSS Feed en-us © 2020 Newgen KnowledgeWorks <![CDATA[Using case-level context to classify cancer pathology reports]]> https://www.researchpad.co/article/elastic_article_7869 Individual electronic health records (EHRs) and clinical reports are often part of a larger sequence—for example, a single patient may generate multiple reports over the trajectory of a disease. In applications such as cancer pathology reports, it is necessary not only to extract information from individual reports, but also to capture aggregate information regarding the entire cancer case based off case-level context from all reports in the sequence. In this paper, we introduce a simple modular add-on for capturing case-level context that is designed to be compatible with most existing deep learning architectures for text classification on individual reports. We test our approach on a corpus of 431,433 cancer pathology reports, and we show that incorporating case-level context significantly boosts classification accuracy across six classification tasks—site, subsite, laterality, histology, behavior, and grade. We expect that with minimal modifications, our add-on can be applied towards a wide range of other clinical text-based tasks.

]]>
<![CDATA[Exploit fully automatic low-level segmented PET data for training high-level deep learning algorithms for the corresponding CT data]]> https://www.researchpad.co/article/5c8823d0d5eed0c484639091

We present an approach for fully automatic urinary bladder segmentation in CT images with artificial neural networks in this study. Automatic medical image analysis has become an invaluable tool in the different treatment stages of diseases. Especially medical image segmentation plays a vital role, since segmentation is often the initial step in an image analysis pipeline. Since deep neural networks have made a large impact on the field of image processing in the past years, we use two different deep learning architectures to segment the urinary bladder. Both of these architectures are based on pre-trained classification networks that are adapted to perform semantic segmentation. Since deep neural networks require a large amount of training data, specifically images and corresponding ground truth labels, we furthermore propose a method to generate such a suitable training data set from Positron Emission Tomography/Computed Tomography image data. This is done by applying thresholding to the Positron Emission Tomography data for obtaining a ground truth and by utilizing data augmentation to enlarge the dataset. In this study, we discuss the influence of data augmentation on the segmentation results, and compare and evaluate the proposed architectures in terms of qualitative and quantitative segmentation performance. The results presented in this study allow concluding that deep neural networks can be considered a promising approach to segment the urinary bladder in CT images.

]]>
<![CDATA[An attention based deep learning model of clinical events in the intensive care unit]]> https://www.researchpad.co/article/5c6dca08d5eed0c48452a6e2

This study trained long short-term memory (LSTM) recurrent neural networks (RNNs) incorporating an attention mechanism to predict daily sepsis, myocardial infarction (MI), and vancomycin antibiotic administration over two week patient ICU courses in the MIMIC-III dataset. These models achieved next-day predictive AUC of 0.876 for sepsis, 0.823 for MI, and 0.833 for vancomycin administration. Attention maps built from these models highlighted those times when input variables most influenced predictions and could provide a degree of interpretability to clinicians. These models appeared to attend to variables that were proxies for clinician decision-making, demonstrating a challenge of using flexible deep learning approaches trained with EHR data to build clinical decision support. While continued development and refinement is needed, we believe that such models could one day prove useful in reducing information overload for ICU physicians by providing needed clinical decision support for a variety of clinically important tasks.

]]>
<![CDATA[DeepDrug3D: Classification of ligand-binding pockets in proteins with a convolutional neural network]]> https://www.researchpad.co/article/5c61e8ebd5eed0c48496f3ee

Comprehensive characterization of ligand-binding sites is invaluable to infer molecular functions of hypothetical proteins, trace evolutionary relationships between proteins, engineer enzymes to achieve a desired substrate specificity, and develop drugs with improved selectivity profiles. These research efforts pose significant challenges owing to the fact that similar pockets are commonly observed across different folds, leading to the high degree of promiscuity of ligand-protein interactions at the system-level. On that account, novel algorithms to accurately classify binding sites are needed. Deep learning is attracting a significant attention due to its successful applications in a wide range of disciplines. In this communication, we present DeepDrug3D, a new approach to characterize and classify binding pockets in proteins with deep learning. It employs a state-of-the-art convolutional neural network in which biomolecular structures are represented as voxels assigned interaction energy-based attributes. The current implementation of DeepDrug3D, trained to detect and classify nucleotide- and heme-binding sites, not only achieves a high accuracy of 95%, but also has the ability to generalize to unseen data as demonstrated for steroid-binding proteins and peptidase enzymes. Interestingly, the analysis of strongly discriminative regions of binding pockets reveals that this high classification accuracy arises from learning the patterns of specific molecular interactions, such as hydrogen bonds, aromatic and hydrophobic contacts. DeepDrug3D is available as an open-source program at https://github.com/pulimeng/DeepDrug3D with the accompanying TOUGH-C1 benchmarking dataset accessible from https://osf.io/enz69/.

]]>
<![CDATA[Effort-aware and just-in-time defect prediction with neural network]]> https://www.researchpad.co/article/5c5df316d5eed0c484580cec

Effort-aware just-in-time (JIT) defect prediction is to rank source code changes based on the likelihood of detects as well as the effort to inspect such changes. Accurate defect prediction algorithms help to find more defects with limited effort. To improve the accuracy of defect prediction, in this paper, we propose a deep learning based approach for effort-aware just-in-time defect prediction. The key idea of the proposed approach is that neural network and deep learning could be exploited to select useful features for defect prediction because they have been proved excellent at selecting useful features for classification and regression. First, we preprocess ten numerical metrics of code changes, and then feed them to a neural network whose output indicates how likely the code change under test contains bugs. Second, we compute the benefit cost ratio for each code change by dividing the likelihood by its size. Finally, we rank code changes according to their benefit cost ratio. Evaluation results on a well-known data set suggest that the proposed approach outperforms the state-of-the-art approaches on each of the subject projects. It improves the average recall and popt by 15.6% and 8.1%, respectively.

]]>
<![CDATA[Utilizing longitudinal microbiome taxonomic profiles to predict food allergy via Long Short-Term Memory networks]]> https://www.researchpad.co/article/5c61e8e9d5eed0c48496f3af

Food allergy is usually difficult to diagnose in early life, and the inability to diagnose patients with atopic diseases at an early age may lead to severe complications. Numerous studies have suggested an association between the infant gut microbiome and development of allergy. In this work, we investigated the capacity of Long Short-Term Memory (LSTM) networks to predict food allergies in early life (0-3 years) from subjects’ longitudinal gut microbiome profiles. Using the DIABIMMUNE dataset, we show an increase in predictive power using our model compared to Hidden Markov Model, Multi-Layer Perceptron Neural Network, Support Vector Machine, Random Forest, and LASSO regression. We further evaluated whether the training of LSTM networks benefits from reduced representations of microbial features. We considered sparse autoencoder for extraction of potential latent representations in addition to standard feature selection procedures based on Minimum Redundancy Maximum Relevance (mRMR) and variance prior to the training of LSTM networks. The comprehensive evaluation reveals that LSTM networks with the mRMR selected features achieve significantly better performance compared to the other tested machine learning models.

]]>
<![CDATA[Deep learning image recognition enables efficient genome editing in zebrafish by automated injections]]> https://www.researchpad.co/article/5c3d011cd5eed0c484038663

One of the most popular techniques in zebrafish research is microinjection. This is a rapid and efficient way to genetically manipulate early developing embryos, and to introduce microbes, chemical compounds, nanoparticles or tracers at larval stages. Here we demonstrate the development of a machine learning software that allows for microinjection at a trained target site in zebrafish eggs at unprecedented speed. The software is based on the open-source deep-learning library Inception v3. In a first step, the software distinguishes wells containing embryos at one-cell stage from wells to be skipped with an accuracy of 93%. A second step was developed to pinpoint the injection site. Deep learning allows to predict this location on average within 42 μm to manually annotated sites. Using a Graphics Processing Unit (GPU), both steps together take less than 100 milliseconds. We first tested our system by injecting a morpholino into the middle of the yolk and found that the automated injection efficiency is as efficient as manual injection (~ 80%). Next, we tested both CRISPR/Cas9 and DNA construct injections into the zygote and obtained a comparable efficiency to that of an experienced experimentalist. Combined with a higher throughput, this results in a higher yield. Hence, the automated injection of CRISPR/Cas9 will allow high-throughput applications to knock out and knock in relevant genes to study their mechanisms or pathways of interest in diverse areas of biomedical research.

]]>
<![CDATA[Distillation of the clinical algorithm improves prognosis by multi-task deep learning in high-risk Neuroblastoma]]> https://www.researchpad.co/article/5c141e9bd5eed0c484d27646

We introduce the CDRP (Concatenated Diagnostic-Relapse Prognostic) architecture for multi-task deep learning that incorporates a clinical algorithm, e.g., a risk stratification schema to improve prognostic profiling. We present the first application to survival prediction in High-Risk (HR) Neuroblastoma from transcriptomics data, a task that studies from the MAQC consortium have shown to remain the hardest among multiple diagnostic and prognostic endpoints predictable from the same dataset. To obtain a more accurate risk stratification needed for appropriate treatment strategies, CDRP combines a first component (CDRP-A) synthesizing a diagnostic task and a second component (CDRP-N) dedicated to one or more prognostic tasks. The approach leverages the advent of semi-supervised deep learning structures that can flexibly integrate multimodal data or internally create multiple processing paths. CDRP-A is an autoencoder trained on gene expression on the HR/non-HR risk stratification by the Children’s Oncology Group, obtaining a 64-node representation in the bottleneck layer. CDRP-N is a multi-task classifier for two prognostic endpoints, i.e., Event-Free Survival (EFS) and Overall Survival (OS). CDRP-A provides the HR embedding input to the CDRP-N shared layer, from which two branches depart to model EFS and OS, respectively. To control for selection bias, CDRP is trained and evaluated using a Data Analysis Protocol (DAP) developed within the MAQC initiative. CDRP was applied on Illumina RNA-Seq of 498 Neuroblastoma patients (HR: 176) from the SEQC study (12,464 Entrez genes) and on Affymetrix Human Exon Array expression profiles (17,450 genes) of 247 primary diagnostic Neuroblastoma of the TARGET NBL cohort. On the SEQC HR patients, CDRP achieves Matthews Correlation Coefficient (MCC) 0.38 for EFS and MCC = 0.19 for OS in external validation, improving over published SEQC models. We show that a CDRP-N embedding is indeed parametrically associated to increasing severity and the embedding can be used to better stratify patients’ survival.

]]>
<![CDATA[Deep learning and artificial intelligence in radiology: Current applications and future directions]]> https://www.researchpad.co/article/5c0ae472d5eed0c484589b6d ]]> <![CDATA[A deep learning approach to automatic detection of early glaucoma from visual fields]]> https://www.researchpad.co/article/5c084192d5eed0c484fca162

Purpose

To investigate the suitability of multi-scale spatial information in 30o visual fields (VF), computed from a Convolutional Neural Network (CNN) classifier, for early-glaucoma vs. control discrimination.

Method

Two data sets of VFs acquired with the OCTOPUS 101 G1 program and the Humphrey Field Analyzer 24–2 pattern were subdivided into control and early-glaucomatous groups, and converted into a new image using a novel voronoi representation to train a custom-designed CNN so to discriminate between control and early-glaucomatous eyes. Saliency maps that highlight what regions of the VF are contributing maximally to the classification decision were computed to provide classification justification. Model fitting was cross-validated and average precision (AP) score performances were computed for our method, Mean Defect (MD), square-root of Loss Variance (sLV), their combination (MD+sLV), and a Neural Network (NN) that does not use convolutional features.

Results

CNN achieved the best AP score (0.874±0.095) across all test folds for one data set compared to others (MD = 0.869±0.064, sLV = 0.775±0.137, MD+sLV = 0.839±0.085, NN = 0.843±0.089) and the third best AP score (0.986 ±0.019) on the other one with slight difference from the other methods (MD = 0.986±0.023, sLV = 0.992±0.016, MD+sLV = 0.987±0.017, NN = 0.985±0.017). In general, CNN consistently led to high AP across different data sets. Qualitatively, computed saliency maps appeared to provide clinically relevant information on the CNN decision for individual VFs.

Conclusion

The proposed CNN offers high classification performance for the discrimination of control and early-glaucoma VFs when compared with standard clinical decision measures. The CNN classification, aided by saliency visualization, may support clinicians in the automatic discrimination of early-glaucomatous and normal VFs.

]]>
<![CDATA[Deep-learning-assisted diagnosis for knee magnetic resonance imaging: Development and retrospective validation of MRNet]]> https://www.researchpad.co/article/5c06f02bd5eed0c484c6d281

Background

Magnetic resonance imaging (MRI) of the knee is the preferred method for diagnosing knee injuries. However, interpretation of knee MRI is time-intensive and subject to diagnostic error and variability. An automated system for interpreting knee MRI could prioritize high-risk patients and assist clinicians in making diagnoses. Deep learning methods, in being able to automatically learn layers of features, are well suited for modeling the complex relationships between medical images and their interpretations. In this study we developed a deep learning model for detecting general abnormalities and specific diagnoses (anterior cruciate ligament [ACL] tears and meniscal tears) on knee MRI exams. We then measured the effect of providing the model’s predictions to clinical experts during interpretation.

Methods and findings

Our dataset consisted of 1,370 knee MRI exams performed at Stanford University Medical Center between January 1, 2001, and December 31, 2012 (mean age 38.0 years; 569 [41.5%] female patients). The majority vote of 3 musculoskeletal radiologists established reference standard labels on an internal validation set of 120 exams. We developed MRNet, a convolutional neural network for classifying MRI series and combined predictions from 3 series per exam using logistic regression. In detecting abnormalities, ACL tears, and meniscal tears, this model achieved area under the receiver operating characteristic curve (AUC) values of 0.937 (95% CI 0.895, 0.980), 0.965 (95% CI 0.938, 0.993), and 0.847 (95% CI 0.780, 0.914), respectively, on the internal validation set. We also obtained a public dataset of 917 exams with sagittal T1-weighted series and labels for ACL injury from Clinical Hospital Centre Rijeka, Croatia. On the external validation set of 183 exams, the MRNet trained on Stanford sagittal T2-weighted series achieved an AUC of 0.824 (95% CI 0.757, 0.892) in the detection of ACL injuries with no additional training, while an MRNet trained on the rest of the external data achieved an AUC of 0.911 (95% CI 0.864, 0.958). We additionally measured the specificity, sensitivity, and accuracy of 9 clinical experts (7 board-certified general radiologists and 2 orthopedic surgeons) on the internal validation set both with and without model assistance. Using a 2-sided Pearson’s chi-squared test with adjustment for multiple comparisons, we found no significant differences between the performance of the model and that of unassisted general radiologists in detecting abnormalities. General radiologists achieved significantly higher sensitivity in detecting ACL tears (p-value = 0.002; q-value = 0.019) and significantly higher specificity in detecting meniscal tears (p-value = 0.003; q-value = 0.019). Using a 1-tailed t test on the change in performance metrics, we found that providing model predictions significantly increased clinical experts’ specificity in identifying ACL tears (p-value < 0.001; q-value = 0.006). The primary limitations of our study include lack of surgical ground truth and the small size of the panel of clinical experts.

Conclusions

Our deep learning model can rapidly generate accurate clinical pathology classifications of knee MRI exams from both internal and external datasets. Moreover, our results support the assertion that deep learning models can improve the performance of clinical experts during medical imaging interpretation. Further research is needed to validate the model prospectively and to determine its utility in the clinical setting.

]]>
<![CDATA[A deep learning model for the detection of both advanced and early glaucoma using fundus photography]]> https://www.researchpad.co/article/5c06f04ed5eed0c484c6d678

Purpose

To build a deep learning model to diagnose glaucoma using fundus photography.

Design

Cross sectional case study Subjects, Participants and Controls: A total of 1,542 photos (786 normal controls, 467 advanced glaucoma and 289 early glaucoma patients) were obtained by fundus photography.

Method

The whole dataset of 1,542 images were split into 754 training, 324 validation and 464 test datasets. These datasets were used to construct simple logistic classification and convolutional neural network using Tensorflow. The same datasets were used to fine tune pre-trained GoogleNet Inception v3 model.

Results

The simple logistic classification model showed a training accuracy of 82.9%, validation accuracy of 79.9% and test accuracy of 77.2%. Convolutional neural network achieved accuracy and area under the receiver operating characteristic curve (AUROC) of 92.2% and 0.98 on the training data, 88.6% and 0.95 on the validation data, and 87.9% and 0.94 on the test data. Transfer-learned GoogleNet Inception v3 model achieved accuracy and AUROC of 99.7% and 0.99 on training data, 87.7% and 0.95 on validation data, and 84.5% and 0.93 on test data.

Conclusion

Both advanced and early glaucoma could be correctly detected via machine learning, using only fundus photographs. Our new model that is trained using convolutional neural network is more efficient for the diagnosis of early glaucoma than previously published models.

]]>
<![CDATA[Deep learning for chest radiograph diagnosis: A retrospective comparison of the CheXNeXt algorithm to practicing radiologists]]> https://www.researchpad.co/article/5bfdb378d5eed0c4845c9bae

Background

Chest radiograph interpretation is critical for the detection of thoracic diseases, including tuberculosis and lung cancer, which affect millions of people worldwide each year. This time-consuming task typically requires expert radiologists to read the images, leading to fatigue-based diagnostic error and lack of diagnostic expertise in areas of the world where radiologists are not available. Recently, deep learning approaches have been able to achieve expert-level performance in medical image interpretation tasks, powered by large network architectures and fueled by the emergence of large labeled datasets. The purpose of this study is to investigate the performance of a deep learning algorithm on the detection of pathologies in chest radiographs compared with practicing radiologists.

Methods and findings

We developed CheXNeXt, a convolutional neural network to concurrently detect the presence of 14 different pathologies, including pneumonia, pleural effusion, pulmonary masses, and nodules in frontal-view chest radiographs. CheXNeXt was trained and internally validated on the ChestX-ray8 dataset, with a held-out validation set consisting of 420 images, sampled to contain at least 50 cases of each of the original pathology labels. On this validation set, the majority vote of a panel of 3 board-certified cardiothoracic specialist radiologists served as reference standard. We compared CheXNeXt’s discriminative performance on the validation set to the performance of 9 radiologists using the area under the receiver operating characteristic curve (AUC). The radiologists included 6 board-certified radiologists (average experience 12 years, range 4–28 years) and 3 senior radiology residents, from 3 academic institutions. We found that CheXNeXt achieved radiologist-level performance on 11 pathologies and did not achieve radiologist-level performance on 3 pathologies. The radiologists achieved statistically significantly higher AUC performance on cardiomegaly, emphysema, and hiatal hernia, with AUCs of 0.888 (95% confidence interval [CI] 0.863–0.910), 0.911 (95% CI 0.866–0.947), and 0.985 (95% CI 0.974–0.991), respectively, whereas CheXNeXt’s AUCs were 0.831 (95% CI 0.790–0.870), 0.704 (95% CI 0.567–0.833), and 0.851 (95% CI 0.785–0.909), respectively. CheXNeXt performed better than radiologists in detecting atelectasis, with an AUC of 0.862 (95% CI 0.825–0.895), statistically significantly higher than radiologists' AUC of 0.808 (95% CI 0.777–0.838); there were no statistically significant differences in AUCs for the other 10 pathologies. The average time to interpret the 420 images in the validation set was substantially longer for the radiologists (240 minutes) than for CheXNeXt (1.5 minutes). The main limitations of our study are that neither CheXNeXt nor the radiologists were permitted to use patient history or review prior examinations and that evaluation was limited to a dataset from a single institution.

Conclusions

In this study, we developed and validated a deep learning algorithm that classified clinically important abnormalities in chest radiographs at a performance level comparable to practicing radiologists. Once tested prospectively in clinical settings, the algorithm could have the potential to expand patient access to chest radiograph diagnostics.

]]>
<![CDATA[Deep mapping gentrification in a large Canadian city using deep learning and Google Street View]]> https://www.researchpad.co/article/5c92b35bd5eed0c4843a3e96

Gentrification is multidimensional and complex, but there is general agreement that visible changes to neighbourhoods are a clear manifestation of the process. Recent advances in computer vision and deep learning provide a unique opportunity to support automated mapping or ‘deep mapping’ of perceptual environmental attributes. We present a Siamese convolutional neural network (SCNN) that automatically detects gentrification-like visual changes in temporal sequences of Google Street View (GSV) images. Our SCNN achieves 95.6% test accuracy and is subsequently applied to GSV sequences at 86110 individual properties over a 9-year period in Ottawa, Canada. We use Kernel Density Estimation (KDE) to produce maps that illustrate where the spatial concentration of visual property improvements was highest within the study area at different times from 2007–2016. We find strong concordance between the mapped SCNN results and the spatial distribution of building permits in the City of Ottawa from 2011 to 2016. Our mapped results confirm those urban areas that are known to be undergoing gentrification as well as revealing areas undergoing gentrification that were previously unknown. Our approach differs from previous works because we examine the atomic unit of gentrification, namely, the individual property, for visual property improvements over time and we rely on KDE to describe regions of high spatial intensity that are indicative of gentrification processes.

]]>
<![CDATA[Testing anxiety in undergraduate medical students and its correlation with different learning approaches]]> https://www.researchpad.co/article/5c92b3c1d5eed0c4843a477d

Objectives

Undergraduate medical students experience a considerable amount of stress and anxiety due to frequent exams. The goal of the present study was to examine the development of exam related anxiety and to test for a correlation between anxiety and learning approaches.

Methods

A whole class of 212 medical students was invited to participate in the study. During their first term, trait anxiety and learning approaches were assessed by use of the state-trait-anxiety inventory (STAI-T) and the approaches-and-study-skills-inventory-for-students (ASSIST), respectively. Acute state anxiety was assessed twice in the course of the second term. To that extent, the STAI-S in combination with measuring salivary cortisol were employed immediately before two oral anatomy exams.

Results

Our most important results were that a surface learning approach correlated significantly with anxiety as a trait and that students with a predominantly strategic approach to learning were the least anxious yet academically most successful.

Conclusion

As surface learners are at risk of being academically less successful and because anxiety is a prerequisite for burn-out, we suggest that medical faculties place particular emphasis on conveying strategies for both, coping with stress and successful learning.

]]>
<![CDATA[Development of an intelligent decision support system for ischemic stroke risk assessment in a population-based electronic health record database]]> https://www.researchpad.co/article/5c92b39cd5eed0c4843a4444

Background

Intelligent decision support systems (IDSS) have been applied to tasks of disease management. Deep neural networks (DNNs) are artificial intelligent techniques to achieve high modeling power. The application of DNNs to large-scale data for estimating stroke risk needs to be assessed and validated. This study aims to apply a DNN for deriving a stroke predictive model using a big electronic health record database.

Methods and results

The Taiwan National Health Insurance Research Database was used to conduct a retrospective population-based study. The database was divided into one development dataset for model training (~70% of total patients for training and ~10% for parameter tuning) and two testing datasets (each ~10%). A total of 11,192,916 claim records from 840,487 patients were used. The primary outcome was defined as any ischemic stroke in inpatient records within 3 years after study enrollment. The DNN was evaluated using the area under the receiver operating characteristic curve (AUC or c-statistic). The development dataset included 672,214 patients (a total of 8,952,000 records) of whom 2,060 patients had stroke events. The mean age of the population was 35.5±20.2 years, with 48.5% men. The model achieved AUC values of 0.920 (95% confidence interval [CI], 0.908–0.932) in testing dataset 1 and 0.925 (95% CI, 0.914–0.937) in testing dataset 2. Under a high sensitivity operating point, the sensitivity and specificity were 92.5% and 79.8% for testing dataset 1; 91.8% and 79.9% for testing dataset 2. Under a high specificity operating point, the sensitivity and specificity were 80.3% and 87.5% for testing dataset 1; 83.7% and 87.5% for testing dataset 2. The DNN model maintained high predictability 5 years after being developed. The model achieved similar performance to other clinical risk assessment scores.

Conclusions

Using a DNN algorithm on this large electronic health record database is capable of obtaining a high performing model for assessment of ischemic stroke risk. Further research is needed to determine whether such a DNN-based IDSS could lead to an improvement in clinical practice.

]]>