Fully automatic prediction for efficacy of photodynamic therapy in clinical port-wine stains treatment: a pilot study

Shengnan Ai, Ping Xue, Chengming Wang, Wenxin Zhang, Juicheng Hsieh, Zhenyu Chen, Bin He, Xiao Zhang, Ning Zhang, Ying Gu
2020 IEEE Access  
In this paper, we report a fully automatic method for the prediction of the treatment efficacy of photodynamic therapy during the clinical treatment in port-wine stains. Histogram of oriented gradients (HOG) features were extracted from optical coherence tomography images. Isolation forest (iForest) was used to build classifier based on these features, achieving a sensitivity of 84% and specificity of 91%. Our dataset consists of 336 PWS lesions of 121 patients. We aim to build a comprehensive
more » ... omputational model for the patients who respond positively to the photodynamic therapy, which could be used to sort and identify patients who respond poorly to photodynamic therapy before treatment and prevent them from unnecessary treatment. INDEX TERMS Anomaly detection, machine learning, optical coherence tomography, port-wine stain, support vector data description.
doi:10.1109/access.2020.2972275 fatcat:a6vx44zc3vggrcj6vmz7ctwdsm