Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802775
Yue Guo, Johan Wrammert, Kavita Singh, A. Kc, K. Bradford, Ashok K. Krishnamurthy
Approximately 3% of births require neonatal resuscitation, which has a direct impact on the immediate survival of these infants. This report proposes an automatic video analysis method for neonatal resuscitation performance evaluation, which helps improve the quality of this procedure. More specifically, we design a deep learning based action model which incorporates motion and spatial information in order to classify neonatal resuscitation actions in videos. First, we use a Convolutional Neural Network to select regions containing infants and only keep those that are motion salient. Second, we extract deep spatial-temporal features to train a linear SVM classifier. Finally, we propose a pair-wise model to ensure consistent classification in consecutive frames. We evaluate the proposed method on a dataset consisting of 17 videos and compare the result against the state-of-the-art method for action classification in videos. To our best knowledge, this work is the first to attempt automatic evaluation of neonatal resuscitation videos and identifies several issues that require further work.
{"title":"Automatic analysis of neonatal video data to evaluate resuscitation prformance","authors":"Yue Guo, Johan Wrammert, Kavita Singh, A. Kc, K. Bradford, Ashok K. Krishnamurthy","doi":"10.1109/ICCABS.2016.7802775","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802775","url":null,"abstract":"Approximately 3% of births require neonatal resuscitation, which has a direct impact on the immediate survival of these infants. This report proposes an automatic video analysis method for neonatal resuscitation performance evaluation, which helps improve the quality of this procedure. More specifically, we design a deep learning based action model which incorporates motion and spatial information in order to classify neonatal resuscitation actions in videos. First, we use a Convolutional Neural Network to select regions containing infants and only keep those that are motion salient. Second, we extract deep spatial-temporal features to train a linear SVM classifier. Finally, we propose a pair-wise model to ensure consistent classification in consecutive frames. We evaluate the proposed method on a dataset consisting of 17 videos and compare the result against the state-of-the-art method for action classification in videos. To our best knowledge, this work is the first to attempt automatic evaluation of neonatal resuscitation videos and identifies several issues that require further work.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131628862","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802777
E. Andersson, Rebecca Hsieh, Howard Szeto, R. Farhoodi, Nurit Haspel, F. Jagodzinski
Predicting how amino acid substitutions affect the stability of a protein has relevance to drug design and may help elucidate the mechanisms of disease-causing protein variants. Unfortunately, wet-lab experiments are time intensive, and to the best of our knowledge there are no efficient computational techniques to asses the effect of multiple mutations. In this work we present a new approach for inferring the effects of single and multiple mutations on a protein's structure. Our rMutant algorithm generates in silico mutants with single or multiple amino acid substitutions. We use a graph-theoretic rigidity analysis approach to compute the distributions of rigid cluster sizes of the wild type and mutant structures which we then analyze to infer the effect of the amino acid substitutions. We successfully predict the effects of multiple mutations for which our previous methods were unsuccessful. We validate the predictions of our computational approach against experimental ΔΔG data. To demonstrate the utility of using rigid cluster size distributions to infer the effects of mutations, we also present a Random Forest Machine Learning approach that relies on rigidity data to predict which residues are critical to the stability of a protein. We predict the destabilizing effects of a single or multiple mutations with over 86% accuracy.
{"title":"Assessing how multiple mutations affect protein stability using rigid cluster size distributions","authors":"E. Andersson, Rebecca Hsieh, Howard Szeto, R. Farhoodi, Nurit Haspel, F. Jagodzinski","doi":"10.1109/ICCABS.2016.7802777","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802777","url":null,"abstract":"Predicting how amino acid substitutions affect the stability of a protein has relevance to drug design and may help elucidate the mechanisms of disease-causing protein variants. Unfortunately, wet-lab experiments are time intensive, and to the best of our knowledge there are no efficient computational techniques to asses the effect of multiple mutations. In this work we present a new approach for inferring the effects of single and multiple mutations on a protein's structure. Our rMutant algorithm generates in silico mutants with single or multiple amino acid substitutions. We use a graph-theoretic rigidity analysis approach to compute the distributions of rigid cluster sizes of the wild type and mutant structures which we then analyze to infer the effect of the amino acid substitutions. We successfully predict the effects of multiple mutations for which our previous methods were unsuccessful. We validate the predictions of our computational approach against experimental ΔΔG data. To demonstrate the utility of using rigid cluster size distributions to infer the effects of mutations, we also present a Random Forest Machine Learning approach that relies on rigidity data to predict which residues are critical to the stability of a protein. We predict the destabilizing effects of a single or multiple mutations with over 86% accuracy.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131744072","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802790
Guangyu Yang, L. Florea
Accurate alternative splicing detection and transcript reconstruction are essential to characterize gene regulation and function and to understand development and disease. However, current methods for extracting splicing variation from RNA-seq data only analyze signals from a single sample, which limits transcript reconstruction and fails to detect a complete set of alternative splicing events. We developed a novel feature selection method, JULiP, that analyzes information across multiple samples to identify alternative splicing variation in the form of splice junctions (introns). It formulates the selection problem as a regularized program, utilizing the latent information from multiple RNA-seq samples to construct an accurate and comprehensive intron set. JULiP is highly accurate, and could detect thousands more introns in any one sample, >30% more than the most sensitive single-sample method, and 10% more introns than in the cumulative set of samples, at higher or comparable precision (>98%). Tested assemblers included Cufflinks, CLASS2, StringTie and FlipFlop, and the multi-sample assembler ISP. JULiP is multi-threaded and parallelized, taking only one minute to analyze up to 100 data sets on a multi-computer cluster, and can easily scale up to allow analyses of hundreds and thousands of RNA-seq samples.
{"title":"JULiP: An efficient model for accurate intron selection from multiple RNA-seq samples","authors":"Guangyu Yang, L. Florea","doi":"10.1109/ICCABS.2016.7802790","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802790","url":null,"abstract":"Accurate alternative splicing detection and transcript reconstruction are essential to characterize gene regulation and function and to understand development and disease. However, current methods for extracting splicing variation from RNA-seq data only analyze signals from a single sample, which limits transcript reconstruction and fails to detect a complete set of alternative splicing events. We developed a novel feature selection method, JULiP, that analyzes information across multiple samples to identify alternative splicing variation in the form of splice junctions (introns). It formulates the selection problem as a regularized program, utilizing the latent information from multiple RNA-seq samples to construct an accurate and comprehensive intron set. JULiP is highly accurate, and could detect thousands more introns in any one sample, >30% more than the most sensitive single-sample method, and 10% more introns than in the cumulative set of samples, at higher or comparable precision (>98%). Tested assemblers included Cufflinks, CLASS2, StringTie and FlipFlop, and the multi-sample assembler ISP. JULiP is multi-threaded and parallelized, taking only one minute to analyze up to 100 data sets on a multi-computer cluster, and can easily scale up to allow analyses of hundreds and thousands of RNA-seq samples.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126198990","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802773
Ankit Agrawal, J. S. Mathias, D. Baker, A. Choudhary
Understanding the prognosis of older adults is a big challenge in healthcare research, especially since very little is known about how different comorbidities interact and influence the prognosis. Recently, a electronic healthcare records dataset of 24 patient attributes from Northwestern Memorial Hospital was used to develop predictive models for five year survival outcome. In this study we analyze the same data for discovering hotspots with respect to five year survival using association rule mining techniques. The goal here is to identify characteristics of patient segments where the five year survival fraction is significantly lower/higher than the survival fraction across the entire dataset. A two-stage post-processing procedure was used to identify non-redundant rules. The resulting rules conform with existing biomedical knowledge and provide interesting insights into prognosis of older adults. Incorporating such information into clinical decision making could advance person-centered healthcare by encouraging optimal use of healthcare services to those patients most likely to benefit.
{"title":"Identifying hotspots in five year survival electronic health records of older adults","authors":"Ankit Agrawal, J. S. Mathias, D. Baker, A. Choudhary","doi":"10.1109/ICCABS.2016.7802773","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802773","url":null,"abstract":"Understanding the prognosis of older adults is a big challenge in healthcare research, especially since very little is known about how different comorbidities interact and influence the prognosis. Recently, a electronic healthcare records dataset of 24 patient attributes from Northwestern Memorial Hospital was used to develop predictive models for five year survival outcome. In this study we analyze the same data for discovering hotspots with respect to five year survival using association rule mining techniques. The goal here is to identify characteristics of patient segments where the five year survival fraction is significantly lower/higher than the survival fraction across the entire dataset. A two-stage post-processing procedure was used to identify non-redundant rules. The resulting rules conform with existing biomedical knowledge and provide interesting insights into prognosis of older adults. Incorporating such information into clinical decision making could advance person-centered healthcare by encouraging optimal use of healthcare services to those patients most likely to benefit.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124925662","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802767
P. H. Souza, J. O. Ferreira, T. M. D. A. Barbosa, A. Rocha
The heart rate (HR) and its variability, known as Heart Rate Variability (HRV), are indispensable measurements for cardiorespiratory monitoring, recognition and quantification of emotions, detection of abnormalities, and heart disease control. In general, the acquisition systems for HR and HRV require a contact area for sensor's installation and positioning, creating restrictions and/or obstructions on user's movements. This paper proposes a noninvasive and noncontact technique for HR and HRV acquisition using a camera. The purposed technique consists in the automatic detection of the user's face and utilization of an Independent Component Analysis (ICA) algorithm to separate the necessary signals to determine the HR and HRV. The experiments have shown more than 95% of similarity between the results of the proposed software (HRVCam) in comparison to the results of the photoplethysmography sensor (PPG).
{"title":"HRVCam: A software for real-time feedback of heart rate and HRV","authors":"P. H. Souza, J. O. Ferreira, T. M. D. A. Barbosa, A. Rocha","doi":"10.1109/ICCABS.2016.7802767","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802767","url":null,"abstract":"The heart rate (HR) and its variability, known as Heart Rate Variability (HRV), are indispensable measurements for cardiorespiratory monitoring, recognition and quantification of emotions, detection of abnormalities, and heart disease control. In general, the acquisition systems for HR and HRV require a contact area for sensor's installation and positioning, creating restrictions and/or obstructions on user's movements. This paper proposes a noninvasive and noncontact technique for HR and HRV acquisition using a camera. The purposed technique consists in the automatic detection of the user's face and utilization of an Independent Component Analysis (ICA) algorithm to separate the necessary signals to determine the HR and HRV. The experiments have shown more than 95% of similarity between the results of the proposed software (HRVCam) in comparison to the results of the photoplethysmography sensor (PPG).","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124958823","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802776
M. R. Amin, S. Skiena, M. Schatz
The quality of the Oxford Nanopores long DNA sequence reads has been, to date, lower than other technologies, causing great interest to develop new algorithms that can make use of the data. So far, alignment methods including LAST, BLAST, BWA-MEM and GraphMap have been used to analyze these sequences. However, each of these tools has significant challenges to use with these data: LAST and BLAST require considerable processing time for high sensitivity, BWA-MEM has the smallest average alignment length, and GraphMap aligns many random strings with moderate accuracy. To address these challenges we developed a new read aligner called NanoBLASTer specifically designed for long nanopore reads. In experiments resequencing the well-studied S. cerevisiae (yeast) and Escherichia coli (E. coli) genomes, we show that our algorithm produces longer alignments with higher overall sensitivity than LAST, BLAST and BWA-MEM. We also show that the runtime of NanoBLASTer is faster than GraphMap, BLAST and BWA-MEM.
{"title":"NanoBLASTer: Fast alignment and characterization of Oxford Nanopore single molecule sequencing reads","authors":"M. R. Amin, S. Skiena, M. Schatz","doi":"10.1109/ICCABS.2016.7802776","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802776","url":null,"abstract":"The quality of the Oxford Nanopores long DNA sequence reads has been, to date, lower than other technologies, causing great interest to develop new algorithms that can make use of the data. So far, alignment methods including LAST, BLAST, BWA-MEM and GraphMap have been used to analyze these sequences. However, each of these tools has significant challenges to use with these data: LAST and BLAST require considerable processing time for high sensitivity, BWA-MEM has the smallest average alignment length, and GraphMap aligns many random strings with moderate accuracy. To address these challenges we developed a new read aligner called NanoBLASTer specifically designed for long nanopore reads. In experiments resequencing the well-studied S. cerevisiae (yeast) and Escherichia coli (E. coli) genomes, we show that our algorithm produces longer alignments with higher overall sensitivity than LAST, BLAST and BWA-MEM. We also show that the runtime of NanoBLASTer is faster than GraphMap, BLAST and BWA-MEM.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123139483","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802778
Mohammed D. Aldhoayan, Leming Zhou
Text classification is an important step in many data analysis procedures. The demand on text classification algorithm is booming due to the increase of the amount of digital data, especially in the healthcare field. A customizable and accurate algorithm is expected to produce positive impact on the efficiency of many data analysis procedures. In this work, we proposed a novel algorithm for accurately classifying data entries in huge text files into several pre-determined categories. We built the algorithm with multiple rules according to text similarity, frequency, and weight. For different classification tasks, the algorithm can be conveniently adjusted to process the corresponding data sets. Data sets related to healthcare cost analysis (hospital discharge summary) and medical classification systems (ICD-9) are used to evaluate the algorithm. When the algorithm is used on the ICD-9 data, the overall accuracy of the algorithm was 100%. After the algorithm was used on 7480 healthcare cost entries, the results were then compared with the ones processed manually by a physician, and the accuracy was between 86%–91.6%, and the difference is from different classification of ambiguous entries, which is hard to determine the correct category even when it is done manually because those entries were documented improperly. This new classification algorithm is 3 to 5 times faster than the manual process on the same data set. Therefore, this customizable and accurate text classification algorithm is effective in saving time compared to the manual classification methods.
{"title":"An accurate and customizable text classification algorithm: Two applications in healthcare","authors":"Mohammed D. Aldhoayan, Leming Zhou","doi":"10.1109/ICCABS.2016.7802778","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802778","url":null,"abstract":"Text classification is an important step in many data analysis procedures. The demand on text classification algorithm is booming due to the increase of the amount of digital data, especially in the healthcare field. A customizable and accurate algorithm is expected to produce positive impact on the efficiency of many data analysis procedures. In this work, we proposed a novel algorithm for accurately classifying data entries in huge text files into several pre-determined categories. We built the algorithm with multiple rules according to text similarity, frequency, and weight. For different classification tasks, the algorithm can be conveniently adjusted to process the corresponding data sets. Data sets related to healthcare cost analysis (hospital discharge summary) and medical classification systems (ICD-9) are used to evaluate the algorithm. When the algorithm is used on the ICD-9 data, the overall accuracy of the algorithm was 100%. After the algorithm was used on 7480 healthcare cost entries, the results were then compared with the ones processed manually by a physician, and the accuracy was between 86%–91.6%, and the difference is from different classification of ambiguous entries, which is hard to determine the correct category even when it is done manually because those entries were documented improperly. This new classification algorithm is 3 to 5 times faster than the manual process on the same data set. Therefore, this customizable and accurate text classification algorithm is effective in saving time compared to the manual classification methods.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132498048","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802768
Wen-Chyi Lin, Ching-Chung Li, J. Epstein, R. Veltri
This paper presents our new result of a study on machine-aided classification of four critical Gleason patterns with curvelet-based texture descriptors extracted from prostatic histological section images. The reliable recognition of these patterns between Gleason score 6 and Gleason score 8 is of crucial importance that will affect the appropriate treatment and patient's quality of life. Higher-order statistical moments of fine scale curvelet coefficients are selected as discriminative features. A two-level classifier consisting of two Gaussian kernel support vector machines, each incorporated with a pertinent voting mechanism by multiple windowed patches in an image for final decision making, has been developed. A set of Tissue MicroArray (TMA) images of four prominent Gleason scores (GS) 3 + 3, 3 + 4, 4 + 3 and 4 + 4 has been studied in machine learning and testing. The testing result has achieved an average accuracy of 93.75% for 4 classes, an outstanding performance when compared with other published works.
{"title":"Curvelet-based texture classification of critical Gleason patterns of prostate histological images","authors":"Wen-Chyi Lin, Ching-Chung Li, J. Epstein, R. Veltri","doi":"10.1109/ICCABS.2016.7802768","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802768","url":null,"abstract":"This paper presents our new result of a study on machine-aided classification of four critical Gleason patterns with curvelet-based texture descriptors extracted from prostatic histological section images. The reliable recognition of these patterns between Gleason score 6 and Gleason score 8 is of crucial importance that will affect the appropriate treatment and patient's quality of life. Higher-order statistical moments of fine scale curvelet coefficients are selected as discriminative features. A two-level classifier consisting of two Gaussian kernel support vector machines, each incorporated with a pertinent voting mechanism by multiple windowed patches in an image for final decision making, has been developed. A set of Tissue MicroArray (TMA) images of four prominent Gleason scores (GS) 3 + 3, 3 + 4, 4 + 3 and 4 + 4 has been studied in machine learning and testing. The testing result has achieved an average accuracy of 93.75% for 4 classes, an outstanding performance when compared with other published works.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114062993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-10-01DOI: 10.1109/ICCABS.2016.7802771
Zhe Xiao, Ruohan Huang, Yi Ding, Tian Lan, Rongfen Dong, Zhiguang Qin, Xinjie Zhang, Wei Wang
Accurate tumor segmentation is an essential and crucial step for computer-aided brain tumor diagnosis and surgical planning. Subjective segmentations are widely adopted in clinical diagnosis and treating, but they are neither accurate nor reliable. An automatical and objective system for brain tumor segmentation is strongly expected. But they are still facing some challenges such as lower segmentation accuracy, demanding a priori knowledge or requiring the human intervention. In this paper, a novel and new coarse-to-fine method is proposed to segment the brain tumor. This hierarchical framework consists of preprocessing, deep learning network based classification and post-processing. The preprocessing is used to extract image patches for each MR image and obtains the gray level sequences of image patches as the input of the deep learning network. The deep learning network based classification is implemented by a stacked auto-encoder network to extract the high level abstract feature from the input, and utilizes the extracted feature to classify image patches. After mapping the classification result to a binary image, the post-processing is implemented by a morphological filter to get the final segmentation result. In order to evaluate the proposed method, the experiment was applied to segment the brain tumor for the real patient dataset. The final performance shows that the proposed brain tumor segmentation method is more accurate and efficient.
{"title":"A deep learning-based segmentation method for brain tumor in MR images","authors":"Zhe Xiao, Ruohan Huang, Yi Ding, Tian Lan, Rongfen Dong, Zhiguang Qin, Xinjie Zhang, Wei Wang","doi":"10.1109/ICCABS.2016.7802771","DOIUrl":"https://doi.org/10.1109/ICCABS.2016.7802771","url":null,"abstract":"Accurate tumor segmentation is an essential and crucial step for computer-aided brain tumor diagnosis and surgical planning. Subjective segmentations are widely adopted in clinical diagnosis and treating, but they are neither accurate nor reliable. An automatical and objective system for brain tumor segmentation is strongly expected. But they are still facing some challenges such as lower segmentation accuracy, demanding a priori knowledge or requiring the human intervention. In this paper, a novel and new coarse-to-fine method is proposed to segment the brain tumor. This hierarchical framework consists of preprocessing, deep learning network based classification and post-processing. The preprocessing is used to extract image patches for each MR image and obtains the gray level sequences of image patches as the input of the deep learning network. The deep learning network based classification is implemented by a stacked auto-encoder network to extract the high level abstract feature from the input, and utilizes the extracted feature to classify image patches. After mapping the classification result to a binary image, the post-processing is implemented by a morphological filter to get the final segmentation result. In order to evaluate the proposed method, the experiment was applied to segment the brain tumor for the real patient dataset. The final performance shows that the proposed brain tumor segmentation method is more accurate and efficient.","PeriodicalId":306466,"journal":{"name":"2016 IEEE 6th International Conference on Computational Advances in Bio and Medical Sciences (ICCABS)","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2016-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125854459","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：