Fig 1.
The block diagram of the work.
Table 1.
Baseline patient and tumor characteristics.
Fig 2.
Imaging appearance and texture analysis features of early stage (a) and advanced stage (b) tonsil cancer.
ROI for tonsil cancer and contralateral normal tonsils are demonstrated. Representative texture images with SSF 0, 3, 5 and coarse filters are shown. The red color in texture images indicates positive pixel values, and the blue color indicates negative pixel values, which are made after filtration according to filters of different sizes. The tumor side is more heterogeneous than the normal side and tends to have more negative pixel values. This feature is more noticeable with the SSF 0, fine filter.
Table 2.
Comparisons of texture analysis-derived parameters between normal tonsil and tonsil cancer.
Table 3.
Comparisons of texture analysis-derived parameters among normal tonsil, early tonsil cancer, and advanced tonsil cancer.
Fig 3.
The ROC curve analysis was performed to differentiate tonsil cancer from normal tonsils.
The AUC for entropy with SSF 0 and 4 was the highest at 0.91.