eduzhai > Applied Sciences > Engineering >

CIDMP Completely Interpretable Detection of Malaria Parasite in Red Blood Cells using Lower-dimensional Feature Space

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 11 pages

Abstract: Predicting if red blood cells (RBC) are infected with the malaria parasite isan important problem in Pathology. Recently, supervised machine learningapproaches have been used for this problem, and they have had reasonablesuccess. In particular, state-of-the-art methods such as Convolutional NeuralNetworks automatically extract increasingly complex feature hierarchies fromthe image pixels. While such generalized automatic feature extraction methodshave significantly reduced the burden of feature engineering in many domains,for niche tasks such as the one we consider in this paper, they result in twomajor problems. First, they use a very large number of features (that may ormay not be relevant) and therefore training such models is computationallyexpensive. Further, more importantly, the large feature-space makes it veryhard to interpret which features are truly important for predictions. Thus, acriticism of such methods is that learning algorithms pose opaque black boxesto its users, in this case, medical experts. The recommendation of suchalgorithms can be understood easily, but the reason for their recommendation isnot clear. This is the problem of non-interpretability of the model, and thebest-performing algorithms are usually the least interpretable. To addressthese issues, in this paper, we propose an approach to extract a very smallnumber of aggregated features that are easy to interpret and compute, andempirically show that we obtain high prediction accuracy even with asignificantly reduced feature-space.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×