We have developed prediction models of tumor growth trajectories (TGTs) based on pretreatment computed tomography (CT) images, prior to targeted therapy with four specific tyrosine kinase inhibitors (TKIs)—erlotinib, gefitinib, afatinib, and osimertinib—for epidermal growth factor receptor (EGFR)-mutated non–small-cell lung carcinoma (NSCLC) patients. TGTs of the time-variant number of tumor cells were predicted for individual patients with tumor growth equations under the assumption that each tumor could contain three cell types: TKI drug-sensitive, drug-persister, and drug-resistant populations. Seven parameters of the TGT models were estimated by support vector regression, which learned the relationships between principal component features and the referenced parameters that were optimized by a Levenberg–Marquardt method. The Spearman correlation coefficient was employed to evaluate the correlation in the number of tumor cells at each CT acquisition timepoint between the reference numbers (derived from CT images) and predicted numbers. The average Spearman correlation was 0.822 (p=0.073) for a training dataset (27 treatments) and 0.818 (p=0.042) for a test dataset (10 treatments). The proposed models could predict TGTs for TKI non-treated patients, thereby estimating how lung tumors will respond to specific TKI drugs to optimize the selection of treatment strategies.