The immunohistochemistry examination is extremely important for identifying antigens in the confirmatory assessment of a patient's diagnosis and prognosis. To successfully perform this task, it is essential to identify specific cells in a given tissue sample correctly. However, carrying out this process manually by a healthcare professional is costly and can lead to errors that affect the patient's health. Thus, for analyzing immunohistochemistry medical images, this work proposes a method based on superpixel segmentation and transport-based morphometry composed by (i) superpixel generation; (ii) color filtering; (iii) cell segmentation; and (iv) transport-based classification. Experimental results show that the proposed method outperforms state-of-the-art methods in the literature for cell classification, needing only 10 positive and 10 negative cell samples to be trained.