Bladder cancer detection by fluorescence imaging with Hexvix: analysis and processing of images obtained during high magnification cystoscopy

Fluorescence cystoscopy has been recently acknowledged as a useful method to detect early superficial bladder cancer, even flat lesions. After the instillation of hexaminolevulinic acid (Hexvix) in the bladder for about an hour, photoactivable porphyrins (PaP), mainly protoporphyrin IX (PpIX) accumulate in the cancerous cells. Although we observe a selective production of PpIX and an outstanding sensitivity of this method, false positive (FP) lesions negatively impact its specificity. Carcinogenesis often combines with angiogenesis, and thus changes in vascular architecture. Therefore, the visualization of the vascular modifications on the fluorescence positive sites is likely to differentiate false and true positive (TP). New methods including high magnification (HM) cystoscopy are being investigated by our group, and will yield a reduced number of biopsies and a better characterization of the fluorescence positive sites. In this study, we are using a dedicated rigid cystoscope, allowing conventional magnification during "macroscopic" observation, as well as image acquisition with HM when the endoscope is in contact with the tissue. Each observed site is biopsied and described by histopathological analysis. The vascular organization (tortuosity, vascular loops, vascular area and diameter) of the fluorescence positive sites was characterized in parallel with an in situ visual grading and a dedicated software procedure. We describe here a simple image processing prototype that classifies the HM images into two classes, according to their pixel distributions. For that purpose, we developed an algorithm in the image spatial and frequency domain, so that the vascular architecture could be described objectively and quantitatively.