Hepatocellular Carcinoma Recognition from Ultrasound Images Through Convolutional Neural Networks and Their Combinations

The Hepatocellular Carcinoma (HCC) represents the most frequent malignant liver tumor. It evolves from cirrhosis after a restructuring phase, at the end of which dysplastic nodules result, which can transform into HCC. The needle biopsy is the golden standard for HCC diagnosis, being, however, invasive, dangerous, as it could lead to infections, respectively to the spread of the tumor through the body. Ultrasonography is a medical examination method which is non-invasive, inexpensive, thus safe, and repeatable. In our research, we developed computerized, non-invasive methods for computer aided and automatic diagnosis of HCC, based on ultrasound images. In the current work, we explored the role of representative Convolutional Neural Networks (CNN), respectively of their combinations, to achieve an optimal classification accuracy. The considered CNNs were fused at classifier level, by employing various combination schemes, based on relevant feature selection, respectively on Kernel Principal Component Analysis (KPCA). At the end, a classification accuracy above 95% resulted.