This paper presents a novel approach for the classification of acute leukemia subtypes using image processing and mathematical techniques. The preprocessing phase analyses 376 features from abdnormal leukocytes images. The features or parameters are Leukemia Parameters that helps to lymphoblastic subtypes detection which come from bone marrow images with heterogeneous staining. The second phase imply the robust generalized principal component analysis as segmentation method for data classification into a subspace arrangement with tree dimensions for each plane of lymphoblastic subtype and four dimension for the subspace arrangement. The novel of our proposal states that the two subtypes of acute leukemia can be classified into a subspace arrangement trough robust generalized principal component analysis method. The subspace arrangement is achieved with singular value decomposition, an hibrid linear model to noise samples detection and homogeneus polynomial. Test reveals that variation in dimension of subspace arrangement depends on features size, the outliers percentage and noise parameters are tunned, dimension of subspace and effective dimension are adjusted, time in execution algorithm and segmentation percentage are measured to lymphoblastic subtypes classification with only 4 parameters from 376 attributes set that are previously computed from cell images and their respective nucleous and cytoplasm.
Digital Object Identifier (DOI)
Flores-Pulido, Leticia; Rodr?guez-G?mez, Gustavo; and A. A. Gonz?lez, Jes?s
"Dimension Reduction Parameters for Leukemia Diagnostic based in Subspace Arrangement Segmentation,"
Applied Mathematics & Information Sciences: Vol. 08:
6, Article 14.
Available at: https://digitalcommons.aaru.edu.jo/amis/vol08/iss6/14