Leveraging persistent homology for liver tumour classification

Distinguishing between intrahepatic cholangiocarcinoma (ICC) and hepatocellular carcinoma (HCC) in imaging is a difficult task for a radiologist. We endeavoured to develop reliable models to automatically classify these tumour types. In this study, we propose to use persistent homology (PH), from the field of topological data analysis (TDA) to build topological shapes from computed tomography (CT) scans of the liver. PH is used to extract topological and geometrical summaries such as the number of persistent connected components and loops from CT scans in the form of persistent barcodes. Topological and geometrical features encapsulated by barcodes are stable to small perturbations to the input data such as variations in scan protocols. Extracted topological features are used as input to various classifiers achieving 97.56% F1-score with 97.5% accuracy. Similar experiment is performed with radiomics features achieving comparable metrics with TDA being marginally higher. Furthermore, pre-trained convolutional neural networks (CNNs) are also explored for benchmarking where comparable results are achieved. Our results suggest that TDA is an effective feature engineering approach for CT scans because, unlike traditional approaches, it uses persistent homology to capture the spatial distribution of texture and pixels in CT scans. Although CNNs achieve comparable results, TDA is preferable due to its interpretability and efficiency. Furthermore, TDA and radiomics features can complement each other whereby a miss-classified scan by radiomics is correctly classified by TDA and vice versa.