Medical image enhancement using split spectrum processing

In ultrasonic medical imaging the echoes from unwanted scatterers manifest themselves as coherent noise superimposed on tissue structure giving the image its granular appearance. In cases where the amplitude of this coherent background noise is comparable to the echo from the abnormality under investigation, ultrasonic detection and characterization of the abnormality becomes very difficult. In this work, the Split Spectrum Processing (SSP) technique is applied to breast and liver sonograms to enhance the detectability of tumorous regions. SSP algorithms successfully utilized in nondestructive testing were modified for medical imaging, based on the scattering and absorption properties of tissue. These algorithms are applied to B-scan images from cancer patients to assess the tumor detection capabilities. Comparison of the experimental results with hospital reports suggest that SSP provides impressive speckle suppression, and is a viable technique for medical image enhancement. SSP also outperformed conventional techniques such as thresholding, logarithmic compression and bandpass filtering. The experimental results reported here are encouraging and justify further work