The aim of the paper is to find the bases for a quantitative study dealing with liver characterization by triphasic CT. This technique allows the acquisition of three abdominal volumes: arterial phase (25 seconds after the contrast injection), venous phase (after 60-90 s), and delayed phase (after 3-5 minutes) [1]. Even though the availability of multitemporal digital volumes represents a significant help to experts for the diagnosis, there is a lack of comprehensive studies devoted to quantitative model characterization and parameter selection which can fully exploit the data information content. The statistical analysis of liver and other tissues is considered an important step in the context of diagnosis and classification [], but detailed studies on multitemporal parameters are only partially proposed. In this work, by referring to several regions of liver parenchyma, a statistical study is applied to various slices in each phase of the training patients cases. Densitometry and texture parameters are extracted and the following features are computed [] from about 1400 regions: 1. First-order statistical parameters from Hounsfield Unit (HU) histogram: mean, median, standard deviation (Sdv), mode, skewness (Sk), kurtosis (Kur); 2. Spatial second-order parameters from gray level co-occurrence matrix (GLCM): entropy (Ent), contrast (Con), energy (En), and homogeneity (Hg). Outliers are found by means of the following formula: x ≤ Q1-1.5(Q3-Q1) (1) x ≥ Q3+1.5(Q3-Q1) (2) where is the parameter value, is the first quartile, and the third quartile. Equation (1) finds the inferior outliers, while equation (2) the superior ones. After deleting the outliers, the average value is computed for each parameter.
“Liver tissue characterization and multitemporal parameter selection in triphasic CT”
Laura, Gemme;Roberta, Ferretti;Silvana , Dellepiane
2015-01-01
Abstract
The aim of the paper is to find the bases for a quantitative study dealing with liver characterization by triphasic CT. This technique allows the acquisition of three abdominal volumes: arterial phase (25 seconds after the contrast injection), venous phase (after 60-90 s), and delayed phase (after 3-5 minutes) [1]. Even though the availability of multitemporal digital volumes represents a significant help to experts for the diagnosis, there is a lack of comprehensive studies devoted to quantitative model characterization and parameter selection which can fully exploit the data information content. The statistical analysis of liver and other tissues is considered an important step in the context of diagnosis and classification [], but detailed studies on multitemporal parameters are only partially proposed. In this work, by referring to several regions of liver parenchyma, a statistical study is applied to various slices in each phase of the training patients cases. Densitometry and texture parameters are extracted and the following features are computed [] from about 1400 regions: 1. First-order statistical parameters from Hounsfield Unit (HU) histogram: mean, median, standard deviation (Sdv), mode, skewness (Sk), kurtosis (Kur); 2. Spatial second-order parameters from gray level co-occurrence matrix (GLCM): entropy (Ent), contrast (Con), energy (En), and homogeneity (Hg). Outliers are found by means of the following formula: x ≤ Q1-1.5(Q3-Q1) (1) x ≥ Q3+1.5(Q3-Q1) (2) where is the parameter value, is the first quartile, and the third quartile. Equation (1) finds the inferior outliers, while equation (2) the superior ones. After deleting the outliers, the average value is computed for each parameter.File | Dimensione | Formato | |
---|---|---|---|
ISBI15GemmeFerrettiDellepiane.pdf
accesso aperto
Descrizione: Articolo
Tipologia:
Documento in Post-print
Dimensione
333.79 kB
Formato
Adobe PDF
|
333.79 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.