After extensive clinical practice, Traditional Chinese Medicine (TCM) tongue diagnosis has gained increasing recognition worldwide. With the advent of artificial intelligence (AI), there is a growing number of intelligent TCM tongue diagnosis technologies that have achieved promising results. Although these methods significantly enhance the efficiency of tongue diagnosis and address the issue of subjectivity, the current approaches based on black-box techniques remain challenging to comprehend. This paper compares various tongue syndrome diagnostic methods and combines various visualization methods to represent the connection between tongue features and syndromes. A total of 831 volunteers diagnosed with non-alcoholic fatty liver disease(NAFLD) were included as the dataset for analysis in this study. Clinical experts from China (chief physicians) categorized the patients into four syndromes: liver depression and spleen deficiency, damp-heat internal accumulation, dampness turbidity internal stagnation, and phlegm-blood stasis mutual obstruction. The experiment presented the tongue features corresponding to each syndrome from a network perspective, along with the regions of interest in tongue diagnosis emphasized by different neural networks. This study has made contributions to the explanatory and objective aspects of tongue diagnosis. As far as we know, this is the first attempt to visualize the relationship between tongue images and syndromes.