The current practice of assessing infants' pain depends on using subjective tools that fail to meet rigorous psychometric standards and requires continuous monitoring by health professionals. Therefore, pain may be misinterpreted or totally missed leading to misdiagnosis and over/under treatment. To address these shortcomings, the current practice can be augmented with a machine-based assessment tool that continuously monitors various pain cues and provides a consistent and minimally biased evaluation of pain. Several machine-based approaches have been proposed to assess infants' pain based on analysis of whether behavioral or physiological pain indictors (i.e., single modality). The aim of this review paper is to provide the reader with the current machine-based approaches in assessing infants' pain. It also proposes the development of a multimodal machine-based pain assessment tool and presents preliminary implementation results.