LCD with Statistical Method

The statistical method is applicable to any homogenous test object. The simple test object used in this work consisted of an acetate sheet with the following details fixed on it with thin tape on the detail borders:

• a central 99.99% pure aluminium platelet (3 x 3 cm2, 0.49 mm thick) to determine LCth in terms of %contrast;

• a 99.99% pure aluminium step wedge (from 0.20 to 1.00 mm in 0.20 mm steps, each step is 0.6 x 0.8 cm2) for correlation of LCth to the absolute physical quantity “aluminium thickness” (mm Al);

• a lead platelet (2 x 2 cm2, 0.5 mm thick) in lateral position for automatic pattern recognition.

A Matlab code for automated analysis of LCD test object images was developed. A square ROI (120 x 120 pixels) was used to evaluate LCth according to the statistical method. To decrease uncertainty, the evaluation was carried out for 9 non overlapping ROIs arranged in the central uniform Al region and the average value LCth considered for each detail size.

In order to express LCth in terms of a more physical quantity, namely threshold Al thickness, the correlation between image contrast and aluminium thickness has been determined. For that purpose, square ROIs of 50 pixels were placed on each step of the Al step wedge and linear regression between mean pixel value and Al thickness calculated for each image.

The Statistical LCD calculation algorithm consists in a pixel-binning process repeated for different pixel binning matrix sizes (2x2, 4x4, etc).

Finally, the so determined data points (LCth) were interpolated to calculate LCth values for detail sizes corresponding to those in the CDMAM phantom or CDRAD phantom.

To further quantify IQ, for each contrast detail curve a summarising figure of merit was used, the Inverse Image Quality Figure IQFinv, calculated as:

IQFinv=1/(∑di*LCth),

where diameter and contrast threshold refer to the contrast detail curve.