Potential of high dimensional radiomic features to assess blood components in intraaortic vessels in non-contrast CT scans

Table 2 Top 20 radiomic features with highest variable importance based on measurement of correlations with hemoglobin and hematocrit values

Features	Hemoglobin p-value	Hematocrit p-value
firstorder-Median	< 0.001	< 0.001
firstorder-Mean	< 0.001	< 0.001
firstorder-RootMeanSquared	< 0.001	< 0.001
firstorder-TotalEnergy	< 0.001	< 0.001
firstorder-90Percentile	< 0.001	< 0.001
firstorder-10Percentile	< 0.001	< 0.001
firstorder-Maximum	< 0.001	< 0.001
firstorder-Energy	0.002	0.001
firstorder-Minimum	0.032	0.014
glszm-GrayLevelNonUniformity	0.052	0.023
glszm-LowGrayLevelZoneEmphasis	0.069	0.074
glcm-MaximumProbability	0.083	0.108
glrlm-ShortRunLowGrayLevelEmphasis	0.101	0.109
glszm-SmallAreaLowGrayLevelEmphasis	0.083	0.115
glcm-Idmn	0.118	0.128
ngtdm-Contrast	0.149	0.135
glszm-SmallAreaEmphasis	0.094	0.138
glrlm-LowGrayLevelRunEmphasis	0.124	0.141
glszm-SmallAreaHighGrayLevelEmphasis	0.162	0.149
gldm-LowGrayLevelEmphasis	0.133	0.15

Measurement of correlation of all radiomic features with hemoglobin and hematocrit levels obtained ± 24 h to the acquisition of the computed tomography images. Measurement of probability used for hypothesis testing is depicted as p-value. Significant values are labeled in bold font. Top 20 features are shown, sorted according to the hematocrit p-value and with the matching hemoglobin p-value

ISSN: 1471-2342