eduzhai > Applied Sciences > Engineering >

An Efficient Confidence Measure-Based Evaluation Metric for Breast Cancer Screening Using Bayesian Neural Networks

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: Screening mammograms is the gold standard for detecting breast cancer early.While a good amount of work has been performed on mammography imageclassification, especially with deep neural networks, there has not been muchexploration into the confidence or uncertainty measurement of theclassification. In this paper, we propose a confidence measure-based evaluationmetric for breast cancer screening. We propose a modular network architecture,where a traditional neural network is used as a feature extractor with transferlearning, followed by a simple Bayesian neural network. Utilizing a two-stageapproach helps reducing the computational complexity, making the proposedframework attractive for wider deployment. We show that by providing themedical practitioners with a tool to tune two hyperparameters of the Bayesianneural network, namely, fraction of sampled number of networks and minimumprobability, the framework can be adapted as needed by the domain expert.Finally, we argue that instead of just a single number such as accuracy, atuple (accuracy, coverage, sampled number of networks, and minimum probability)can be utilized as an evaluation metric of our framework. We provideexperimental results on the CBIS-DDSM dataset, where we show the trends inaccuracy-coverage tradeoff while tuning the two hyperparameters. We also showthat our confidence tuning results in increased accuracy with a reduced set ofimages with high confidence when compared to the baseline transfer learning. Tomake the proposed framework readily deployable, we provide (anonymized) sourcecode with reproducible results at this https URL.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×