Tumor Detection Improvement in Scintimammographic imaging: Multivariate Image Analysis Approach

The use of dedicated gamma camera, with high intrinsic spatial resolution can raise the detection of sub centimeter cancers in Scintimammography. The recent development of new gamma imagers based on scintillation array with high spatial resolution has strongly unproved the SNR of the resulting images; however, Compton scattering contamination is the main drawback in Scintimammography, since it limits the sensitivity of tumor detection, especially in the portion of the breast close to the chest. In this paper we introduce the Multivariate Image Analysis as a numerical technique to improve the tumor detection and tissue identification in scintimammographic imaging. The total pulse height distribution resulting from the detector is considered as series of images viewing the same field-of-view. The stack of images resulting from different intervals of the photon spectra is a multivariate image and can be studied by the principal image component analysis. This method is shown as a two-step procedure: calculation of loadings and calculation of scores. Scores can be shown visually and by the global pixel correlation and anticorrelation the contribution of chest, healthy tissues, and tumor can be highlighted. The results show the realistic possibility to depict scintimammographic images as a series of principal components images that are able to separate Compton contamination, improving the visibility of breast regions with higher Tc 99m Sestamibi uptake