Association between RhoB protein expression and rectal cancer in radiotherapy (RT) resistance has been hypothesized. However, there is no strong clinical evidence to confirm the prognostic power of the protein in this disease. Here, we combine advanced artificial intelligence and signal processing methods to examine RhoB expression captured by immunohistochemical imaging of tumor tissue in a cohort of rectal cancer patients with preoperative RT. Prediction results obtained from the proposed approach with 10-fold cross-validation accuracy rates between 85% and 94% not only discover the potential role of RhoB for rectal cancer prognosis, but also significantly outperform individual pretrained deep-learning models with accuracy between 58% and 67%.