Classification of Sonoelastography Images of Prostate Cancer Using Transformation-Based Feature Extraction Techniques

Abstract Recent trends in medical imaging enable overcoming the limitations of noninvasive histological assessment and subjectivity concerns in diagnosis of several diseases. These advancements are complemented by computer-assisted diagnosis (CAD) algorithms to further enhance accuracy of diagnosis. CAD algorithms are able to monitor the prognosis of chronic diseases like cancer. Among different cancer types, prostate cancer is the most common cancer in men. Early diagnosis of this cancer is known to significantly enhance the survival rate of patients. In this chapter, a CAD algorithm has been presented for sonoelastography images, which is an ultrasound-based imaging technique that determines cancers through analyzing tissue elasticity. Fourier and wavelet-based transformations were used to classify the images after segmenting them into affected and unaffected regions based on the ground truth information obtained from clinician. Classification rate of a maximum of 97% was obtained using the stationary wavelet transform (SWT) method, whereas 94%, 96%, and 61% were obtained with fast Fourier transform (FFT), discrete wavelet transform (DWT), and spatial domain, respectively. The proposed study shows that sonoelastography images can be used for diagnosis of prostate cancer by applying wavelet technique.