Liver cancer (hepatocellular carcinoma (HCC)) poses a significant global health burden, and an accurate diagnosis relies on whole slide image (WSI) analysis by pathologists. While WSI analysis is the gold standard, challenges arise due to the massive size and complexity of WSIs. Deep learning offers promise in computational pathology, but some limitations exist including computational cost and the need for expensive and time-consuming WSI annotation. This research addresses these limitations by proposing an approach using Mask R-CNN on WSI patches. Our model aims to generate probability heatmaps for specific liver cancer tissues which will help us do multi-class spatial distribution analysis. Our research will contribute to the list of models that can be used for liver WSI classification. It will provide automatic WSI classification, increase interpretability, and potentially aid pathologists in accurate liver cancer diagnosis and prognosis. The successful development of this approach will contribute to liver cancer prognosis and improved clinical workflows.