Skip to main content

Table 3 MaZda classification results – results in groups of T2-weighted images.

From: Non-Hodgkin lymphoma response evaluation with MRI texture classification

T2-weighted images classification

RDA

PCA

LDA

NDA

Examinations

Feature selection method

mis%

mis%

mis%

mis%

E1, E2, E3

Combination E1, E2, E3

34%

35%

47%

30%

E1, E2

Combination E1, E2, E3

29%

29%

39%

19%

 

Combination E1, E2

37%

35%

40%

35%

E1, E3

Combination E1, E2, E3

15%

14%

19%

4%

 

Combination E1, E3

16%

17%

21%

4%

E2, E3

Combination E1, E2, E3

25%

24%

25%

14%

 

Combination E2, E3

24%

23%

30%

12%

  1. Imaging timepoint (E1, E2, E3) combinations for classification analyses. Feature selection methods given in rows. Misclassification percentage (mis%) given for raw data analysis (RDA), principal component analysis (PCA), linear discriminant analysis (LDA) and non-linear discriminant analysis (NDA) in columns. "Combination E1, E2, E3" in feature selection methods refers to features, which have proved to give best discrimination in all imaging timepoints analyses with Fisher and POE+ACC methods, combination of two imaging timepoints refers respectively to features from the analyses in question.