An image analysis method for prostate tissue classification: preliminary validation with resonance sensor data

Abstract

Resonance sensor systems have been shown to be able to distinguish between cancerous and normal prostate tissue, in vitro. The aim of this study was to improve the accuracy of the tissue determination, to simplify the tissue classification process with computerized morphometrical analysis, to decrease the risk of human errors, and to reduce the processing time. In this article we present our newly developed computerized classification method based on image analysis. In relation to earlier resonance sensor studies we increased the number of normal prostate tissue classes into stroma, epithelial tissue, lumen and stones. The linearity between the impression depth and tissue classes was calculated using multiple linear regression (R² = 0.68, n = 109, p < 0.001) and partial least squares (R² = 0.55, n = 109, p < 0.001). Thus it can be concluded that there existed a linear relationship between the impression depth and the tissue classes. The new image analysis method was easy to handle and decreased the classification time by 80%.

Keywords:

• Prostate cancer
• Stiffness
• Partial least square
• Image analysis
• Resonance sensor