The Performance Evaluation of a Distributed Image Classification Pipeline Based on Hadoop and MapReduce with Initial Application to Medical Images
The amount of medical image data acquired in hospitals is increasing very rapidly, which makes the analysis of medical imaging data a big challenge. One of the trends is to use big data technology to distribute computation loads, but the distributed implementation of various medical analysis algorithms and their performance are not fully investigated. This paper uses Hadoop clusters and the MapReduce paradigm to implement a medical classification pipeline including the feature extraction, the K-means clustering and a back propagation neural network with additional momentum. The focus is to analyze the performance impacts of different factors and compare them with the single node case. The experiments turned out that for the combination of learning rate 0.7 and the momentum factor 0.9 can achieve a good compromise between the prediction accuracy and training time. The performance evaluation also includes the relationship between the training time and size of training data, between the training time and the number of iterations, between the prediction accuracy and the number of iterations, and etc. The proposed pipeline is successfully applied to the actual classification of breast cancer images.
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
No Article Media
Document Type: Research Article
Publication date: January 1, 2018
More about this publication?
- Journal of Medical Imaging and Health Informatics (JMIHI) is a medium to disseminate novel experimental and theoretical research results in the field of biomedicine, biology, clinical, rehabilitation engineering, medical image processing, bio-computing, D2H2, and other health related areas.
- Editorial Board
- Information for Authors
- Subscribe to this Title
- Ingenta Connect is not responsible for the content or availability of external websites