A pattern analysis approach is proposed to model spatial interaction between the immune cells and the tumor islets within the tumor microenvironment. Embedded in a statistical null hypothesis paradigm, the proposed solution provides a novel mechanism to extract quantitative, system-level snapshot of the tumor microenvironment during immunotherapy. Experimental results on synthetic and real histopathological data suggest the potential of our technique in mining robust informatics from the spatial distribution of immune cells at different time-points during cancer treatment, and this could be potentially adopted in large-scale experimental studies in the future to objectively monitor disease progression and the patient's response to immune therapy.