Компьютерная оптика (Aug 2022)

Improving the efficiency of brain MRI image analysis using feature selection

  • V.V. Konevsky,
  • A.V. Blagov,
  • A.V. Gaidel,
  • A.V. Kapishnikov,
  • A.V. Kupriyanov,
  • E.N. Surovtsev,
  • D.G. Asatryan

DOI
https://doi.org/10.18287/2412-6179-CO-1040
Journal volume & issue
Vol. 46, no. 4
pp. 621 – 627

Abstract

Read online

This article discusses the possibility of improving the quality of analysis of MRI images of the brain in various scanning modes by using greedy feature selection algorithms. A total of five MRI sequences were reviewed. The texture features were formed using the MaZda software package. Using an algorithm for recursive feature selection, the accuracy of determining the type of tumor can be increased from 69% to 100%. With the help of the combined algorithm for the selection of signs, it was possible to increase the accuracy of determining the need for treatment of a patient from 60% to 75% and from 81% to 88% in the case of using an additional class of data for patients whose accurate result of treatment is unknown. The use of textural features in combination with a feature that is responsible for the type of meningioma made it possible to unambiguously determine the need for patient treatment.

Keywords