Parsed Region Feature Extraction Method for Hepatocellular Carcinoma Detection Using Medical Images

Hepatocellular carcinoma detection using medical images relies on precise feature selection from the set of extracted ones that augments high detection accuracy. This article introduces a Parsed Region Feature Extraction Method (PRFEM) to improve the carcinoma detection accuracy from medial input images. The proposed method utilizes a two-layer convoluted neural network with a linear pooling function to select optimal features for detection. The first layer is responsible for selecting visual pattern-based features that improve the region detection. In the second layer, the identified features are verified for their missing distribution sequences that are useful in retaining precision. The linear pooling identifies the distribution break points to train the CNN with the need for high feature selection. This improves the detection precision with minimum errors. The proposed PRFEM improves detection precision, feature selection, and accuracy by 11.7%, 9.4%, and 10.5% respectively for the varying patterns.