Recurrence prediction with local binary pattern-based dosiomics in patients with head and neck squamous cell carcinoma

We investigated an approach for predicting recurrence after radiation therapy using local binary pattern (LBP)-based dosiomics in patients with head and neck squamous cell carcinoma (HNSCC). Recurrence/non-recurrence data were collected from 131 patients after intensity-modulated radiation therapy. The cases were divided into training (80%) and test (20%) datasets. A total of 327 dosiomics features, including cold spot volume, first-order features, and texture features, were extracted from the original dose distribution (ODD) and LBP on gross tumor volume, clinical target volume, and planning target volume. The CoxNet algorithm was employed in the training dataset for feature selection and dosiomics signature construction. Based on a dosiomics score (DS)-based Cox proportional hazard model, two recurrence prediction models (DS_ODD and DS_LBP) were constructed using the ODD and LBP dosiomics features. These models were used to evaluate the overall adequacy of the recurrence prediction using the concordance index (CI), and the prediction performance was assessed based on the accuracy and area under the receiver operating characteristic curve (AUC). The CIs for the test dataset were 0.71 and 0.76 for DS_ODD and DS_LBP, respectively. The accuracy and AUC for the test dataset were 0.71 and 0.76 for the DS_ODD model and 0.79 and 0.81 for the DS_LBP model, respectively. LBP-based dosiomics models may be more accurate in predicting recurrence after radiation therapy in patients with HNSCC.