Determining mutations in the kinase domain of the epidermal growth factor receptor (EGFR) is critical for the effectiveness of EGFR tyrosine kinase inhibitors (TKIs) in lung cancer. Yet, DNA-based sequencing analysis of tumor samples is time-consuming and only provides gene mutation information on EGFR, making it challenging to design effective EGFR-TKI therapeutic strategies. Here, we present a new image-based method involving the rational design of a quenched probe based on EGFR-TKI to identify mutant proteins, which permits specific and "no-wash" real-time imaging of EGFR in living cells only upon covalent targeting of the EGFR kinase. We also show that the probe enables distinguishing EGFR mutant tumor-bearing mice from wild-type tumor-bearing mice via fluorescence-intensity-based imaging with high signal contrast. More interestingly, the image-based phenotypic approach can be used to predict EGFR mutations in tumors from lung cancer patients with an accuracy of 94%. Notably, when immunohistochemistry analysis is integrated, an improved accuracy of 98% is achieved. These data delineate a drug-based phenotypic imaging approach for in-biopsy visualization and define functional groups of EGFR mutants that can effectively guide EGFR-TKI therapeutic decision-making besides gene mutation analysis.