Figure 3
From: Image-based Classification of Tumor Type and Growth Rate using Machine Learning: a preclinical study
Importance of features in tumor classification. (a) Feature importance was estimated for the random forest model using the entire tumor volume and a GLCM of size 512. The top 10 most important features included both 1st and 2nd order texture features. (b) Cluster prominence derived from a GLCM of size 512 was calculated for each tumor type. Cluster prominence was significantly difference between medulloblastoma and glioma, both U87 and GL261 tumors. Box plot represents 25th and 75th percentiles; bar = min and max values; line = median, ***p < 0.0002, ****p < 0.0001.
