Detection of fibrosis in liver biopsy images using multi-objective genetic programming

Abstract

This paper proposes an automatic construction of feature extractor for liver fibrosis detection using a multiobjective genetic programming approach in which a constructed feature extractor was measured in different aspects in which becomes the objectives of the evolutionary run. The result of the evolutionary run is a set of solutions with different strengths and weaknesses. A solution from each experiment is selected and compared with a benchmark handcraft method in by each experiment and top-five manners. One of the best result obtained has 2.09 fibrosis estimation error which is less than the benchmark method with 2.63 fibrosis estimation error.