Development of a Prototype for Hepatocellular Carcinoma Classification Based on Morphological Features Automatically Measured in Whole Slide Images

Abstract

The advent of new digital imaging technologies including high-throughput slide scanners is making a very compelling case as part of the clinical workflow. Tools developed for morphometric image analysis are accelerating the transition of pathology into a more quantitative science. The system for detection of regions suspected to be cancerous in gastric and colorectal tissue is already available. There is a real need for not only cancer detection but also quantification of histological features, because quantitative morphological characteristics can include important diagnostic and prognostic information. If an association between quantitative features and clinical findings is indicated, quantification of morphological features would be extremely useful to select the best treatment. Image measurement technology also has the potential for investigative pathology. We have developed a prototype system for both quantification of morphological features and automated identification of hepatocellular carcinoma (HCC) within whole slide images (WSI) of liver biopsy based on image recognition and measurement techniques. Our system displays quantified cell and tissue features as histogram, bar graph, and heat map on the screen. Displaying all features in such a unified visualization makes it easy to interpret quantitative feature. In this paper, we present a prototype designed specifically for morphological feature visualization in an easy-to-understand manner.