LAB AUTOMATIC EVALUATION OF COLON CLEANSING

Aims:

For objective estimation of colon cleansing, we propose a LAB color selection procedure to automatically detect and compute the Bowel Preparation Coefficient (BPC) on more scales, on the informative frames of a video colonoscopy.

Methods:

The advantage of LAB color space is that one of the three numerical values used to express color, is the lightness L*. The other two axes green-red (a*) and blue-yellow (b*) components, are more approached to the human perception than other color spaces.

To retain only the relevant video colonoscopy frames and to decrease the complexity, we detect the edges on each frame, we test the entropy levels on sub-frames and we discard the non-informative images. On a semi-supervised detection we identify the color cube and color squares corresponding to residual material presence on bowels, slightly varying for each patient. Areas for valid frames are computed and summed, obtaining the overall final estimation.

Results:

Tests have been made on tens of thousands of frames. We may obtain up to 50.000 frames from a 20 minutes video colonoscopy. We used 17 videos of slightly different lengths. Blurred frames, highly lighted, light reflections were automatically discarded. On the remaining frames we identified fecal residues and computed a*b* color features. For speed, we take into consideration only the two color variables during the final total color evaluation on the valid frames of a video.

Conclusions:

Previous attempts have been made using RGB color space, characterized by computing burden. Our method offers an algorithm easy to compute, results being obtained faster. This procedure can be applied to the video recordings, saving time and facilitating the computer-assisted analysis of the cleansing aspects, relevant for diagnosis in colonoscopy.