scholarly journals Detecting abnormal fundus images by employing deep transfer learning

2019 ◽  
Author(s):  
yan yu ◽  
Changfan Wu ◽  
Xiao Chen ◽  
Xiangbing Zhu ◽  
Yinfen Hou ◽  
...  
Keyword(s):  
Transfer Learning ◽  
Learning Algorithm ◽  
Care Center ◽  
High Sensitivity ◽  
Vascular Lesion ◽  
Fundus Photography ◽  
Fundus Images ◽  
Validation Data ◽  
Data Set ◽  
Health Care Center

Abstract BackgroundTo develop and validate a deep transfer learning (DTL) algorithm in detecting abnormalities of fundus images from non-mydriatic fundus photography examination.Methods1,295 fundus images from January 2017 to December 2018 at Yijishan Hospital of Wannan Medical College were collected for developing and validating the deep transfer learning algorithm in detecting abnormal fundus images. The DTL model was developed by using 929(normal 254, abnormal 402) fundus images, including normal fundus images and abnormal fundus images, the latter including, maculopathy, optic neuropathy, vascular lesion, choroidal lesions, vitreous disease, cataract and the others. We tested our model using a subset of the publically available MESSIDOR dataset (using 366 images) and evaluate the testing performance of the DTL model for detecting abnormal fundus images. ResultsIn the internal validation data set (n=273 images), the AUC, sensitivity, accuracy and specificity of the DTL for correctly classified funds images were 0.997, 97.41%, 97.07% and 96.82%, respectively. For test data set (n=273 images), the AUC, sensitivity, accuracy and specificity of the DTL for correctly classification funds images were 0.926, 88.17%, 87.18% and 86.67%, respectively.ConclusionIn the evaluation, the DTL presented high sensitivity and specificity for detecting abnormal fundus-related diseases. Further research is necessary to improve this method and evaluate the applicability of the DTL in the community health care center.

scholarly journals Detecting abnormal fundus images by employing deep transfer learning

2020 ◽  
Author(s):  
Yan Yu ◽  
Xiao Chen ◽  
Xiang-Bing Zhu ◽  
Peng-Fei Zhang ◽  
Yin-Fen Hou ◽  
...  
Keyword(s):  
Transfer Learning ◽  
Learning Algorithm ◽  
Care Center ◽  
High Sensitivity ◽  
Vascular Lesion ◽  
Fundus Photography ◽  
Validation Dataset ◽  
Fundus Images ◽  
Internal Validation ◽  
Health Care Center

Abstract Background: To develop and validate a deep transfer learning (DTL) algorithm for detecting abnormalities in fundus images from nonmydriatic fundus photography examinations. Methods : A total of 1,295 fundus images from January 2017 to December 2018 at Yijishan Hospital of Wannan Medical College were collected for developing and validating the deep transfer learning algorithm in detecting abnormal fundus images. The DTL model was developed by using 929 (normal 254, abnormal 402) fundus images, including normal fundus images and abnormal fundus images, the latter including maculopathy, optic neuropathy, vascular lesion, choroidal lesions, vitreous disease, and cataracts. We tested our model using a subset of the publicly available Messidor dataset (using 366 images) and evaluated the testing performance of the DTL model for detecting abnormal fundus images. Results : In the internal validation dataset (n=273 images), the AUC, sensitivity, accuracy, and specificity of the DTL for correctly classified fundus images were 0.997, 97.41%, 97.07%, and 96.82%, respectively. For the test dataset (n=273 images), the AUC, sensitivity, accuracy, and specificity of the DTL for correctly classifying fundus images were 0.926, 88.17%, 87.18%, and 86.67%, respectively. Conclusion : In the evaluation, the DTL presented high sensitivity and specificity for detecting abnormal fundus-related diseases. Further research is necessary to improve this method and evaluate the applicability of the DTL in the community health care center. Key words : Fundus images; Deep transfer learning; Developing and validation; Artificial intelligence.

Event detection of different English data sources based on transfer learning

2021 ◽  
pp. 1-11
Author(s):  
Yanan Huang ◽  
Yuji Miao ◽  
Zhenjing Da
Keyword(s):  
Transfer Learning ◽  
Event Detection ◽  
Visual Analysis ◽  
Learning Algorithm ◽  
Data Sources ◽  
Data Set ◽  
Data Source ◽  
Single Data Source ◽  
The Difference ◽  
Single Data

The methods of multi-modal English event detection under a single data source and isomorphic event detection of different English data sources based on transfer learning still need to be improved. In order to improve the efficiency of English and data source time detection, based on the transfer learning algorithm, this paper proposes multi-modal event detection under a single data source and isomorphic event detection based on transfer learning for different data sources. Moreover, by stacking multiple classification models, this paper makes each feature merge with each other, and conducts confrontation training through the difference between the two classifiers to further make the distribution of different source data similar. In addition, in order to verify the algorithm proposed in this paper, a multi-source English event detection data set is collected through a data collection method. Finally, this paper uses the data set to verify the method proposed in this paper and compare it with the current most mainstream transfer learning methods. Through experimental analysis, convergence analysis, visual analysis and parameter evaluation, the effectiveness of the algorithm proposed in this paper is demonstrated.

Design of novel multi filter union feature selection framework for breast cancer dataset

2021 ◽  
pp. 1063293X2110160
Author(s):  
Dinesh Morkonda Gunasekaran ◽  
Prabha Dhandayudam
Keyword(s):  
Breast Cancer ◽  
Feature Selection ◽  
Care Center ◽  
Feature Selection Method ◽  
Selection Method ◽  
Cancer Center ◽  
Breast Cancer Dataset ◽  
Data Set ◽  
Health Care Center ◽  
Cancer Data

Nowadays women are commonly diagnosed with breast cancer. Feature based Selection method plays an important step while constructing a classification based framework. We have proposed Multi filter union (MFU) feature selection method for breast cancer data set. The feature selection process based on random forest algorithm and Logistic regression (LG) algorithm based union model is used for selecting important features in the dataset. The performance of the data analysis is evaluated using optimal features subset from selected dataset. The experiments are computed with data set of Wisconsin diagnostic breast cancer center and next the real data set from women health care center. The result of the proposed approach shows high performance and efficient when comparing with existing feature selection algorithms.

scholarly journals A Transfer Learning Approach for Malignant Prostate Lesion Detection on Multiparametric MRI

2019 ◽  
Vol 18 ◽  
pp. 153303381985836 ◽  
Author(s):  
Quan Chen ◽  
Shiliang Hu ◽  
Peiran Long ◽  
Fang Lu ◽  
Yujie Shi ◽  
...  
Keyword(s):  
Magnetic Resonance Imaging ◽  
Magnetic Resonance ◽  
Transfer Learning ◽  
Area Under The Curve ◽  
Learning Approach ◽  
Resonance Imaging ◽  
Validation Data ◽  
Data Set ◽  
Noncancerous Tissue ◽  
Multiparametric Magnetic Resonance Imaging

Purpose: In prostate focal therapy, it is important to accurately localize malignant lesions in order to increase biological effect of the tumor region while achieving a reduction in dose to noncancerous tissue. In this work, we proposed a transfer learning–based deep learning approach, for classification of prostate lesions in multiparametric magnetic resonance imaging images. Methods: Magnetic resonance imaging images were preprocessed to remove bias artifact and normalize the data. Two state-of-the-art deep convolutional neural network models, InceptionV3 and VGG-16, were pretrained on ImageNet data set and retuned on the multiparametric magnetic resonance imaging data set. As lesion appearances differ by the prostate zone that it resides in, separate models were trained. Ensembling was performed on each prostate zone to improve area under the curve. In addition, the predictions from lesions on each prostate zone were scaled separately to increase the area under the curve for all lesions combined. Results: The models were tuned to produce the highest area under the curve on validation data set. When it was applied to the unseen test data set, the transferred InceptionV3 model achieved an area under the curve of 0.81 and the transferred VGG-16 model achieved an area under the curve of 0.83. This was the third best score among the 72 methods from 33 participating groups in ProstateX competition. Conclusion: The transfer learning approach is a promising method for prostate cancer detection on multiparametric magnetic resonance imaging images. Features learned from ImageNet data set can be useful for medical images.

Towards an Essential Data Set: Applicability in the Domain of Maternal Health Services

1992 ◽  
Vol 31 (03) ◽  
pp. 182-192 ◽  
Author(s):  
A. K. Singh ◽  
K. Boström ◽  
S. Chowdhury ◽  
E. Trell ◽  
O. Wigertz ◽  
...  
Keyword(s):  
Health Care ◽  
Information System ◽  
Maternal Health ◽  
Health Services ◽  
Care Center ◽  
Maternal Health Services ◽  
Care Organization ◽  
Process Of Care ◽  
Data Set ◽  
Health Care Center

Abstract:There is a need for consensus on the quantity of data that must be available in a computer-based information system of a health care organization. In this paper we take up the issue of defining the data content of an information system and introduce the concept of Essential Data Sets with an explicit methodology which was applied to define a data set for the Maternal Health Services program. A key step in the method was a recognized technique used in systems development process called data modelling, in this case infological modelling, by an interdisciplinary group. A preliminary set of 86 data elements was identified and it provided the foundation for development of an application software for discussion and a real-world testing framework. The acceptability of the data set was tested in a laboratory perspective by retrospective data entry from records of 94 pregnant women registered at a maternal health care center in Sweden. Data from a total of 1,318 prenatal visits, an outcome visit, and a postnatal visit for each woman was entered into a computer using the software, with no loss of information. Thus, in a short-term perspective the acceptability of the data set was demonstrated. The software has since been implemented for pilot prospective studies at sites in India and Sweden. The use of a common data protocol is an essential foundation for patient outcome research, especially as the trend of health care management has changed from a “process of care” orientation to an “outcome of care” orientation.

scholarly journals Adaptive Transfer Learning Based on a Two-Stream Densely Connected Residual Shrinkage Network for Transformer Fault Diagnosis over Vibration Signals

Electronics ◽  
2021 ◽  
Vol 10 (17) ◽  
pp. 2130
Author(s):  
Xiaoyan Liu ◽  
Yigang He ◽  
Lei Wang
Keyword(s):  
Deep Learning ◽  
Fault Diagnosis ◽  
Transfer Learning ◽  
Learning Algorithm ◽  
Operating Conditions ◽  
Data Set ◽  
Vibration Signals ◽  
Time Frequency ◽  
Diagnosis Method ◽  
Transformer Fault

Vibration signal analysis is an efficient online transformer fault diagnosis method for improving the stability and safety of power systems. Operation in harsh interference environments and the lack of fault samples are the most challenging aspects of transformer fault diagnosis. High-precision performance is difficult to achieve when using conventional fault diagnosis methods. Thus, this study proposes a transformer fault diagnosis method based on the adaptive transfer learning of a two-stream densely connected residual shrinkage network over vibration signals. First, novel time-frequency analysis methods (i.e., Synchrosqueezed Wavelet Transform and Synchrosqueezed Generalized S-transform) are proposed to convert vibration signals into different images, effectively expanding the samples and extracting effective features of signals. Second, a Two-stream Densely Connected Residual Shrinkage (TSDen2NetRS) network is presented to achieve a high accuracy fault diagnosis under different working conditions. Furthermore, the Residual Shrinkage layer (RS layer) is applied as a nonlinear transformation layer to the deep learning framework to remove unimportant features and enhance anti-interference performance. Lastly, an adaptive transfer learning algorithm that can automatically select the source data set by using the domain measurement method is proposed. This algorithm accelerates the training of the deep learning network and improves accuracy when the number of samples is small. Vibration experiments of transformers are conducted under different operating conditions, and their results show the effectiveness and robustness of the proposed method.

scholarly journals Biomarker Discovery for Meta-Classification of Melanoma Metastatic Progression using Transfer Learning

2021 ◽  
Author(s):  
Jose Marie Antonio Miñoza ◽  
Jonathan Adam Rico ◽  
Pia Regina Fatima Zamora ◽  
Manny Bacolod ◽  
Reinhard Laubenbacher ◽  
...  
Keyword(s):  
Transfer Learning ◽  
Biomarker Discovery ◽  
The Cancer Genome Atlas ◽  
Metastatic Progression ◽  
Validation Data ◽  
Data Set ◽  
Diagnosis And Prognosis ◽  
Ensemble Machine Learning ◽  
Machine Learning Model ◽  
Cancer Genome Atlas

Melanoma is considered the most serious and aggressive type of skin cancer, and metastasis appears to be the most important factor in prognosticating this type of cancer. With the emergence of new therapeutic strategies for metastatic melanoma that have shown improvement in patient survival, we developed a transfer learning-based biomarker discovery model that could help in the diagnosis and prognosis of this disease. After applying it to the ensemble machine learning model, results reveal that the genes we found show consistency with other methodologies previously applied to the same TCGA (The Cancer Genome Atlas) data set, and our methods found novel biomarker genes as well. Our ensemble model achieved Area Under the Receiver Operating Characteristic (AUC) of 0.9861, an accuracy of 91.05, and an F1 score of 90.60 using an independent validation data set. This study was able to identify potential genes for diagnostic classification (C7 and GRIK5) and diagnostic and prognostic biomarkers (S100A7, S100A7, KRT14, KRT17, KRT6B, KRTDAP, SERPINB4, TSHR, PVRL4, WFDC5, IL20RB). We also assessed the potential sources of bias for our model and confirmed some of them by the model's performance.

scholarly journals Development and Validation of a Deep Learning Algorithm and Open-Source Platform for the Automatic Labelling of Motion Capture Markers

2021 ◽  
Author(s):  
Allison L. Clouthier ◽  
Gwyneth B. Ross ◽  
Matthew P. Mavor ◽  
Isabel Coll ◽  
Alistair Boyle ◽  
...  
Keyword(s):  
Neural Network ◽  
Deep Learning ◽  
Open Source ◽  
Transfer Learning ◽  
Motion Capture ◽  
Learning Algorithm ◽  
Simulated Data ◽  
Motion Capture Data ◽  
Commercial Software ◽  
Data Set

AbstractThe purpose of this work was to develop an open-source deep learning-based algorithm for motion capture marker labelling that can be trained on measured or simulated marker trajectories. In the proposed algorithm, a deep neural network including recurrent layers is trained on measured or simulated marker trajectories. Labels are assigned to markers using the Hungarian algorithm and a predefined generic marker set is used to identify and correct mislabeled markers. The algorithm was first trained and tested on measured motion capture data. Then, the algorithm was trained on simulated trajectories and tested on data that included movements not contained in the simulated data set. The ability to improve accuracy using transfer learning to update the neural network weights based on labelled motion capture data was assessed. The effect of occluded and extraneous markers on labelling accuracy was also examined. Labelling accuracy was 99.6% when trained on measured data and 92.8% when trained on simulated trajectories, but could be improved to up to 98.8% through transfer learning. Missing or extraneous markers reduced labelling accuracy, but results were comparable to commercial software. The proposed labelling algorithm can be used to accurately label motion capture data in the presence of missing and extraneous markers and accuracy can be improved as data are collected, labelled, and added to the training set. The algorithm and user interface can reduce the time and manual effort required to label optical motion capture data, particularly for those with limited access to commercial software.

scholarly journals Novel Machine Learning Approaches for Modelling the Gully Erosion Susceptibility

2020 ◽  
Vol 12 (17) ◽  
pp. 2833 ◽  
Author(s):  
Alireza Arabameri ◽  
Omid Asadi Nalivan ◽  
Subodh Chandra Pal ◽  
Rabin Chakrabortty ◽  
Asish Saha ◽  
...  
Keyword(s):  
Machine Learning ◽  
Water Conservation ◽  
Learning Algorithm ◽  
Gully Erosion ◽  
Support Vector ◽  
Learning Approaches ◽  
Operating Characteristics ◽  
Validation Data ◽  
Data Set ◽  
Jackknife Test

The extreme form of land degradation caused by the formation of gullies is a major challenge for the sustainability of land resources. This problem is more vulnerable in the arid and semi-arid environment and associated damage to agriculture and allied economic activities. Appropriate modeling of such erosion is therefore needed with optimum accuracy for estimating vulnerable regions and taking appropriate initiatives. The Golestan Dam has faced an acute problem of gully erosion over the last decade and has adversely affected society. Here, the artificial neural network (ANN), general linear model (GLM), maximum entropy (MaxEnt), and support vector machine (SVM) machine learning algorithm with 90/10, 80/20, 70/30, 60/40, and 50/50 random partitioning of training and validation samples was selected purposively for estimating the gully erosion susceptibility. The main objective of this work was to predict the susceptible zone with the maximum possible accuracy. For this purpose, random partitioning approaches were implemented. For this purpose, 20 gully erosion conditioning factors were considered for predicting the susceptible areas by considering the multi-collinearity test. The variance inflation factor (VIF) and tolerance (TOL) limit were considered for multi-collinearity assessment for reducing the error of the models and increase the efficiency of the outcome. The ANN with 50/50 random partitioning of the sample is the most optimal model in this analysis. The area under curve (AUC) values of receiver operating characteristics (ROC) in ANN (50/50) for the training and validation data are 0.918 and 0.868, respectively. The importance of the causative factors was estimated with the help of the Jackknife test, which reveals that the most important factor is the topography position index (TPI). Apart from this, the prioritization of all predicted models was estimated taking into account the training and validation data set, which should help future researchers to select models from this perspective. This type of outcome should help planners and local stakeholders to implement appropriate land and water conservation measures.

scholarly journals DeepOMe: A Web Server for the Prediction of 2′-O-Me Sites Based on the Hybrid CNN and BLSTM Architecture

2021 ◽  
Vol 9 ◽  
Author(s):  
Hongyu Li ◽  
Li Chen ◽  
Zaoli Huang ◽  
Xiaotong Luo ◽  
Huiqin Li ◽  
...  
Keyword(s):  
High Performance ◽  
Large Scale ◽  
Short Term Memory ◽  
Learning Algorithm ◽  
Web Server ◽  
High Sensitivity ◽  
Computational Method ◽  
Regulatory Control ◽  
Data Set ◽  
Deep Learning Algorithm

2′-O-methylations (2′-O-Me or Nm) are one of the most important layers of regulatory control over gene expression. With increasing attentions focused on the characteristics, mechanisms and influences of 2′-O-Me, a revolutionary technique termed Nm-seq were established, allowing the identification of precise 2′-O-Me sites in RNA sequences with high sensitivity. However, as the costs and complexities involved with this new method, the large-scale detection and in-depth study of 2′-O-Me is still largely limited. Therefore, the development of a novel computational method to identify 2′-O-Me sites with adequate reliability is urgently needed at the current stage. To address the above issue, we proposed a hybrid deep-learning algorithm named DeepOMe that combined Convolutional Neural Networks (CNN) and Bidirectional Long Short-term Memory (BLSTM) to accurately predict 2′-O-Me sites in human transcriptome. Validating under 4-, 6-, 8-, and 10-fold cross-validation, we confirmed that our proposed model achieved a high performance (AUC close to 0.998 and AUPR close to 0.880). When testing in the independent data set, DeepOMe was substantially superior to NmSEER V2.0. To facilitate the usage of DeepOMe, a user-friendly web-server was constructed, which can be freely accessed at http://deepome.renlab.org.

Sign in / Sign up

Export Citation Format

Share Document