The feasibility of personalized medicine for cancer treatment is largely hampered by costly, labor-intensive and time-consuming models for drug discovery. Herein, establishing new pre-clinical models to tackle these issues for personalized medicine is urgently demanded. Methods: We established a three-dimensional tumor slice culture (3D-TSC) platform incorporating label-free techniques for time-course experiments to predict anti-cancer drug efficacy and validated the 3D-TSC model by multiphoton fluorescence microscopy, RNA sequence analysis, histochemical and histological analysis. Results: Using time-lapse imaging of the apoptotic reporter sensor C3 (C3), we performed cell-based high-throughput drug screening and shortlisted high-efficacy drugs to screen murine and human 3D-TSCs, which validate effective candidates within 7 days of surgery. Histological and RNA sequence analyses demonstrated that 3D-TSCs accurately preserved immune components of the original tumor, which enables the successful achievement of immune checkpoint blockade assays with antibodies against PD-1 and/or PD-L1. Label-free multiphoton fluorescence imaging revealed that 3D-TSCs exhibit lipofuscin autofluorescence features in the time-course monitoring of drug response and efficacy. Conclusion: This technology accelerates precision anti-cancer therapy by providing a cheap, fast, and easy platform for anti-cancer drug discovery.