Tomy Varghese

My primary research interests are in signal, image processing and machine learning applications in medical ultrasound imaging for the early detection of cancer. Early detection of cancers can significantly improve patient survival rate and lead to effective treatment. Aims of our research are to develop signal-processing and machine learning approaches for extracting relevant tissue microstructure information embedded in the spectra of RF broadband ultrasonic echoes. The ultimate goal is to provide fast, reliable and well-understood signal processing techniques for the early detection of malignant tissue using ultrasound. A unique aspect of this research is its focus on characterizing Fourier phase and magnitude of RF ultrasonic echoes with spectral redundancy characterized by the spectral autocorrelation function (also called the Generalized Spectrum). Disruptions in the normal structure of liver and breast tissue due to invading cancers, are detected by examining the changes in the quasi-periodic nature of the liver tissue and the presence of specular scatterers (calcifications in malignant tissue) in breast tissue using this innovative approach. This method is more accurate than the traditional techniques used to estimate the scatterer spacing.

My current research deals with the development of Elastography. Elastography is rapidly developing into a new ultrasonic imaging modality, capable of producing images of internal strain or tissue elasticity. Elastography is a method for imaging the elastic properties of compliant tissues, producing gray scale strain images referred to as elastograms. You can think of Elastography as a “high tech” form of palpation. Elastography has been used for imaging and characterizing tumors in the breast, prostate, kidney, liver, muscle and other tissues. Ultrasonic visualization is widely used in practically all-medical specialties. The proposed technique could therefore have a large impact on medical practice in the United States. Elastography has also been implemented using MRI, Optical Coherence Tomography (OCT) and other imaging modalities.