With functional information, 18FDG-PET outcome prediction before the start of radiotherapy (RT) provides information for potential dose painting to guide individualized RT design. This work proposes a model of Dose-Distribution-Driven PET Image-based Outcome Prediction (DDD-PIOP), which predicts 18FDG-PET image outcome of oropharyngeal cancer IMRT using pre-treatment images and planned dose distribution. DDD-PIOP centralizes a modified convolutional neural network (CNN) as a deep learning method and a harmonic rectification for image tuning. CNN was constructed with 8 convolutional layers, and each layer was followed by a rectified linear unit activation and batch normalization. Output images from CNN were processed by a harmonic rectification to fine-tune standardized uptake value (SUV) intensity values. DDD-PIOP was demonstrated in a cohort of 66 oropharyngeal cancer patients undergoing 70Gy IMRT. Among 66 patients, 61 were used for DDD-PIOP training and the remaining 5 were used for model testing. Each patient received Pre-RT PET-CT scan and a 2nd PET-CT scan (Intra-RT) after 24Gy delivered in 2Gy/daily fx. The Intra-RT scan was deformed to the Pre-RT scan, and the dose distribution was resampled to the PET-CT image grid size. During the model training, axial slices of Pre-RT PET/CT images, binary masks of GTV/CTV and dose distribution were stacked as the inputs. The loss function was a modified mean square error equation to heavily penalize results in CTV region. 10-fold validation regime was implemented. The output PET images from CNN were processed by a linear rectification function with 3 customized thresholds to harmonize intensity distributions. To evaluate prediction accuracy, mean SUV values in GTV/CTV and gamma passing rate of the predicted PET images in reference to the ground-truth Intra-RT PET images were reported in 5 test patients. DDD-PIOP successfully generated Intra-RT PET image outcome predictions. In test results, DDD-PIOP prediction agreed with ground-truth Intra-RT PET images in general contrasts (verified by radiation oncologists). 3D mean SUV values in CTV/GTV by prediction were 1.36/3.01, which were close to ground truth values (1.47/2.91). In gamma analysis (SUV threshold 10% local value, 3 mm distance-to-agreement), the average 2D gamma passing rates were 85.8% for all 266 axial slices containing CTV structure and 98.8% for all 107 axial slices containing GTV structure. For high uptake axial slices (GTV mean SUV > 10 in Pre-RT PET), the average 2D gamma passing rate was improved to 99.4%. Average 3D gamma passing rates in CTV/GTV were 98.8%/99.9%, respectively. For oropharyngeal cancer patients undergoing definitive IMRT, DDD-PIOP successfully predicted Intra-RT PET image outcome with good quantitative accuracy. Current results demonstrate the potential use of DDD-PIOP in RT planning decision making in future.