Dataset Information

Distillation of crop models to learn plant physiology theories using machine learning.

ABSTRACT: Convolutional neural networks (CNNs) can not only classify images but can also generate key features, e.g., the Google neural network that learned to identify cats by simply watching YouTube videos, for the classification. In this paper, crop models are distilled by CNN to evaluate the ability of deep learning to identify the plant physiology knowledge behind such crop models simply by learning. Due to difficulty in collecting big data on crop growth, a crop model was used to generate datasets. The generated datasets were fed into CNN for distillation of the crop model. The models trained by CNN were evaluated by the visualization of saliency maps. In this study, three saliency maps were calculated using all datasets (case 1) and using datasets with spikelet sterility due to either high temperature at anthesis (case 2) or cool summer damage (case 3). The results of case 1 indicated that CNN determined the developmental index of paddy rice, which was implemented in the crop model, simply by learning. Moreover, CNN identified the important individual environmental factors affecting the grain yield. Although CNN had no prior knowledge of spikelet sterility, cases 2 and 3 indicated that CNN realized about paddy rice becoming sensitive to daily mean and maximum temperatures during specific periods. Such deep learning approaches can be used to accelerate the understanding of crop models and make the models more portable. Moreover, the results indicated that CNN can be used to develop new plant physiology theories simply by learning.

SUBMITTER: Yamamoto K

PROVIDER: S-EPMC6541271 | biostudies-literature | 2019

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Distillation of crop models to learn plant physiology theories using machine learning.

Yamamoto Kyosuke K

PloS one 20190529 5

Convolutional neural networks (CNNs) can not only classify images but can also generate key features, e.g., the Google neural network that learned to identify cats by simply watching YouTube videos, for the classification. In this paper, crop models are distilled by CNN to evaluate the ability of deep learning to identify the plant physiology knowledge behind such crop models simply by learning. Due to difficulty in collecting big data on crop growth, a crop model was used to generate datasets. ...[more]

PMID: 31141528

Similar Datasets

Project description:BACKGROUND AND PURPOSE:MR imaging-based modeling of tumor cell density can substantially improve targeted treatment of glioblastoma. Unfortunately, interpatient variability limits the predictive ability of many modeling approaches. We present a transfer learning method that generates individualized patient models, grounded in the wealth of population data, while also detecting and adjusting for interpatient variabilities based on each patient's own histologic data. MATERIALS AND METHODS:We recruited patients with primary glioblastoma undergoing image-guided biopsies and preoperative imaging, including contrast-enhanced MR imaging, dynamic susceptibility contrast MR imaging, and diffusion tensor imaging. We calculated relative cerebral blood volume from DSC-MR imaging and mean diffusivity and fractional anisotropy from DTI. Following image coregistration, we assessed tumor cell density for each biopsy and identified corresponding localized MR imaging measurements. We then explored a range of univariate and multivariate predictive models of tumor cell density based on MR imaging measurements in a generalized one-model-fits-all approach. We then implemented both univariate and multivariate individualized transfer learning predictive models, which harness the available population-level data but allow individual variability in their predictions. Finally, we compared Pearson correlation coefficients and mean absolute error between the individualized transfer learning and generalized one-model-fits-all models. RESULTS:Tumor cell density significantly correlated with relative CBV (r = 0.33, P < .001), and T1-weighted postcontrast (r = 0.36, P < .001) on univariate analysis after correcting for multiple comparisons. With single-variable modeling (using relative CBV), transfer learning increased predictive performance (r = 0.53, mean absolute error = 15.19%) compared with one-model-fits-all (r = 0.27, mean absolute error = 17.79%). With multivariate modeling, transfer learning further improved performance (r = 0.88, mean absolute error = 5.66%) compared with one-model-fits-all (r = 0.39, mean absolute error = 16.55%). CONCLUSIONS:Transfer learning significantly improves predictive modeling performance for quantifying tumor cell density in glioblastoma.

Project description:Abstract The goal of diagnosing the coronavirus disease 2019 (COVID‐19) from suspected pneumonia cases, that is, recognizing COVID‐19 from chest X‐ray or computed tomography (CT) images, is to improve diagnostic accuracy, leading to faster intervention. The most important and challenging problem here is to design an effective and robust diagnosis model. To this end, there are three challenges to overcome: (1) The lack of training samples limits the success of existing deep‐learning‐based methods. (2) Many public COVID‐19 data sets contain only a few images without fine‐grained labels. (3) Due to the explosive growth of suspected cases, it is urgent and important to diagnose not only COVID‐19 cases but also the cases of other types of pneumonia that are similar to the symptoms of COVID‐19. To address these issues, we propose a novel framework called Unsupervised Meta‐Learning with Self‐Knowledge Distillation to address the problem of differentiating COVID‐19 from pneumonia cases. During training, our model cannot use any true labels and aims to gain the ability of learning to learn by itself. In particular, we first present a deep diagnosis model based on a relation network to capture and memorize the relation among different images. Second, to enhance the performance of our model, we design a self‐knowledge distillation mechanism that distills knowledge within our model itself. Our network is divided into several parts, and the knowledge in the deeper parts is squeezed into the shallow ones. The final results are derived from our model by learning to compare the features of images. Experimental results demonstrate that our approach achieves significantly higher performance than other state‐of‐the‐art methods. Moreover, we construct a new COVID‐19 pneumonia data set based on text mining, consisting of 2696 COVID‐19 images (347 X‐ray + 2349 CT), 10,155 images (9661 X‐ray + 494 CT) about other types of pneumonia, and the fine‐grained labels of all. Our data set considers not only a bacterial infection or viral infection which causes pneumonia but also a viral infection derived from the influenza virus or coronavirus.

Project description:To phenotype mechanistic differences between heart failure with reduced (HFrEF) and preserved (HFpEF) ejection fraction, a closed-loop model of the cardiovascular system coupled with patient-specific transthoracic echocardiography (TTE) and right heart catheterization (RHC) data was used to identify key parameters representing haemodynamics. Thirty-one patient records (10 HFrEF, 21 HFpEF) were obtained from the Cardiovascular Health Improvement Project database at the University of Michigan. Model simulations were tuned to match RHC and TTE pressure, volume, and cardiac output measurements in each patient. The underlying physiological model parameters were plotted against model-based norms and compared between HFrEF and HFpEF. Our results confirm the main mechanistic parameter driving HFrEF is reduced left ventricular (LV) contractility, whereas HFpEF exhibits a heterogeneous phenotype. Conducting principal component analysis, k -means clustering, and hierarchical clustering on the optimized parameters reveal (i) a group of HFrEF-like HFpEF patients (HFpEF1), (ii) a classic HFpEF group (HFpEF2), and (iii) a group of HFpEF patients that do not consistently cluster (NCC). These subgroups cannot be distinguished from the clinical data alone. Increased LV active contractility ( p<0.001 ) and LV passive stiffness ( p<0.001 ) at rest are observed when comparing HFpEF2 to HFpEF1. Analysing the clinical data of each subgroup reveals that elevated systolic and diastolic LV volumes seen in both HFrEF and HFpEF1 may be used as a biomarker to identify HFrEF-like HFpEF patients. These results suggest that modelling of the cardiovascular system and optimizing to standard clinical data can designate subgroups of HFpEF as separate phenotypes, possibly elucidating patient-specific treatment strategies. KEY POINTS: Analysis of data from right heart catheterization (RHC) and transthoracic echocardiography (TTE) of heart failure (HF) patients using a closed-loop model of the cardiovascular system identifies key parameters representing haemodynamic cardiovascular function in patients with heart failure with reduced and preserved ejection fraction (HFrEF and HFpEF). Analysing optimized parameters representing cardiovascular function using machine learning shows mechanistic differences between HFpEF groups that are not seen analysing clinical data alone. HFpEF groups presented here can be subdivided into three subgroups: HFpEF1 described as 'HFrEF-like HFpEF', HFpEF2 as 'classic HFpEF', and a third group of HFpEF patients that do not consistently cluster. Focusing purely on cardiac function consistently captures the underlying dysfunction in HFrEF, whereas HFpEF is better characterized by dysfunction in the entire cardiovascular system. Our methodology reveals that elevated left ventricular systolic and diastolic volumes are potential biomarkers for identifying HFrEF-like HFpEF patients.

Project description:BackgroundAccurate and reliable predictions of infectious disease can be valuable to public health organizations that plan interventions to decrease or prevent disease transmission. A great variety of models have been developed for this task. However, for different data series, the performance of these models varies. Hepatitis E, as an acute liver disease, has been a major public health problem. Which model is more appropriate for predicting the incidence of hepatitis E? In this paper, three different methods are used and the performance of the three methods is compared.MethodsAutoregressive integrated moving average(ARIMA), support vector machine(SVM) and long short-term memory(LSTM) recurrent neural network were adopted and compared. ARIMA was implemented by python with the help of statsmodels. SVM was accomplished by matlab with libSVM library. LSTM was designed by ourselves with Keras, a deep learning library. To tackle the problem of overfitting caused by limited training samples, we adopted dropout and regularization strategies in our LSTM model. Experimental data were obtained from the monthly incidence and cases number of hepatitis E from January 2005 to December 2017 in Shandong province, China. We selected data from July 2015 to December 2017 to validate the models, and the rest was taken as training set. Three metrics were applied to compare the performance of models, including root mean square error(RMSE), mean absolute percentage error(MAPE) and mean absolute error(MAE).ResultsBy analyzing data, we took ARIMA(1, 1, 1), ARIMA(3, 1, 2) as monthly incidence prediction model and cases number prediction model, respectively. Cross-validation and grid search were used to optimize parameters of SVM. Penalty coefficient C and kernel function parameter g were set 8, 0.125 for incidence prediction, and 22, 0.01 for cases number prediction. LSTM has 4 nodes. Dropout and L2 regularization parameters were set 0.15, 0.001, respectively. By the metrics of RMSE, we obtained 0.022, 0.0204, 0.01 for incidence prediction, using ARIMA, SVM and LSTM. And we obtained 22.25, 20.0368, 11.75 for cases number prediction, using three models. For MAPE metrics, the results were 23.5%, 21.7%, 15.08%, and 23.6%, 21.44%, 13.6%, for incidence prediction and cases number prediction, respectively. For MAE metrics, the results were 0.018, 0.0167, 0.011 and 18.003, 16.5815, 9.984, for incidence prediction and cases number prediction, respectively.ConclusionsComparing ARIMA, SVM and LSTM, we found that nonlinear models(SVM, LSTM) outperform linear models(ARIMA). LSTM obtained the best performance in all three metrics of RSME, MAPE, MAE. Hence, LSTM is the most suitable for predicting hepatitis E monthly incidence and cases number.

Project description:BACKGROUND:Diabetes Mellitus is an increasingly prevalent chronic disease characterized by the body's inability to metabolize glucose. The objective of this study was to build an effective predictive model with high sensitivity and selectivity to better identify Canadian patients at risk of having Diabetes Mellitus based on patient demographic data and the laboratory results during their visits to medical facilities. METHODS:Using the most recent records of 13,309 Canadian patients aged between 18 and 90 years, along with their laboratory information (age, sex, fasting blood glucose, body mass index, high-density lipoprotein, triglycerides, blood pressure, and low-density lipoprotein), we built predictive models using Logistic Regression and Gradient Boosting Machine (GBM) techniques. The area under the receiver operating characteristic curve (AROC) was used to evaluate the discriminatory capability of these models. We used the adjusted threshold method and the class weight method to improve sensitivity - the proportion of Diabetes Mellitus patients correctly predicted by the model. We also compared these models to other learning machine techniques such as Decision Tree and Random Forest. RESULTS:The AROC for the proposed GBM model is 84.7% with a sensitivity of 71.6% and the AROC for the proposed Logistic Regression model is 84.0% with a sensitivity of 73.4%. The GBM and Logistic Regression models perform better than the Random Forest and Decision Tree models. CONCLUSIONS:The ability of our model to predict patients with Diabetes using some commonly used lab results is high with satisfactory sensitivity. These models can be built into an online computer program to help physicians in predicting patients with future occurrence of diabetes and providing necessary preventive interventions. The model is developed and validated on the Canadian population which is more specific and powerful to apply on Canadian patients than existing models developed from US or other populations. Fasting blood glucose, body mass index, high-density lipoprotein, and triglycerides were the most important predictors in these models.

Dataset Information

Distillation of crop models to learn plant physiology theories using machine learning.

Publications

Distillation of crop models to learn plant physiology theories using machine learning.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets