Triphasic-aided Liver Lesion Segmentation in Non-contrast CT

According to the latest WHO Globocan in 2020 and the most updated national cancer registry in Egypt in 2014, liver cancer ranks first as the most common cancer representing 20-35% of all cancer prevalence. Due to its high burden and cumulative risk to the population, various campaigns for the treatment of hepatitis virus were started and sealed to mitigate its effect. In screening and diagnosis for liver lesions, contrast agents are mostly used to make abnormalities more visible due to the low contrast differentiation between lesions and liver tissue. Contrast agents supply was heavily affected by COVID-19 global supply chain disruption and local economic instability afterward. Our goal is to establish a benchmark for liver tumor segmentation on non-contrast CT scans and show the potential of utilizing multi-phase data to enhance the training process, thereby enhancing lesion detection accuracy in NC imaging when access to contrast agents is restricted. Our challenge distinguishes itself with multi-phase CT (non-contrast and contrast-enhanced arterial (ART), portal venous (PV), and delayed phase cuts) intending to increase the upper bound for liver lesion segmentation.