Cone-beam CT radiomic signature with AI discriminates architectural phenotypes of jaw lesions with AUC 0.92

A CBCT-based radiomic signature with logistic regression achieved an AUC of 0.92 for differentiating architectural phenotypes of jaw cysts and tumors, with texture features reflecting spatial heterogeneity proving most discriminative.

• Forty radiomic features differed significantly among phenotypes after false discovery rate correction; texture descriptors reflecting gray-level non-uniformity and entropy were most discriminative.
• Complex heterogeneous lesions showed elevated texture heterogeneity compared with homogeneous fluid-dominant lesions, supporting biological relevance.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.