SampleBoost for Capsule Endoscopy Categorization and Abnormality Detection

Analyzing Capsule Endoscopy videos is an expensive process that requires considerable human effort and time. The massive amount of data limits the usage of ensemble learning methods. In this paper SampleBoost, a boosting method that employs novel intelligent sampling, is proposed to learn from capsule endoscopy data. Sample Boost intelligently selects a subset of the training set at each iteration and evens imbalanced classes Experimental results show a great improvement in both accuracy and efficiency as well as avoidance of early termination for both the balanced images categorization and the imbalanced abnormality detection.