Variational image segmentation for endoscopic human colonic aberrant crypt foci.

Area of Science:

• Medical Imaging
• Computational Pathology
• Gastroenterology

Background:

• Aberrant crypt foci (ACF) are recognized precursors to colorectal cancer.
• Early detection of ACF is crucial for clinical intervention and cancer prevention.
• In vivo endoscopic imaging provides a direct view of the colonic mucosa.

Purpose of the Study:

• To introduce a novel variational image segmentation method for assessing ACF in endoscopic colon images.
• To enhance the Chan-Vese active contours model for ACF-specific structural features.
• To enable accurate identification and boundary delineation of ACF.

Main Methods:

• Utilized a variational active contours without edges model (Chan-Vese).
• Employed level sets for representing segmentation boundaries.
• Discretized the model in space using finite elements and in time using finite differences.
• Applied the method to in vivo endoscopic images of the human colon.

Main Results:

• Successfully identified aberrant crypt foci (ACF) in endoscopic images.
• Accurately delineated the boundaries of the detected ACF.
• Visualized internal crypt orifices within the identified ACF structures.

Conclusions:

• The proposed variational image segmentation method effectively assesses ACF in vivo.
• The enhanced active contours model accurately segments ACF and their features.
• This technique holds potential for early colorectal cancer detection and diagnosis.