eduzhai > Applied Sciences > Engineering >

An Aggregate Method for Thorax Diseases Classification

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 15 pages

Abstract: A common problem found in real-word medical image classification is theinherent imbalance of the positive and negative patterns in the dataset wherepositive patterns are usually rare. Moreover, in the classification of multipleclasses with neural network, a training pattern is treated as a positivepattern in one output node and negative in all the remaining output nodes. Inthis paper, the weights of a training pattern in the loss function are designedbased not only on the number of the training patterns in the class but also onthe different nodes where one of them treats this training pattern as positiveand the others treat it as negative. We propose a combined approach of weightscalculation algorithm for deep network training and the training optimizationfrom the state-of-the-art deep network architecture for thorax diseasesclassification problem. Experimental results on the Chest X-Ray image datasetdemonstrate that this new weighting scheme improves classificationperformances, also the training optimization from the EfficientNet improves theperformance furthermore. We compare the aggregate method with severalperformances from the previous study of thorax diseases classifications toprovide the fair comparisons against the proposed method.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...