Computational imaging modalities replace the bulky, complex, and expensive optical components of traditional imaging procedures with numerical reconstruction steps. Digital holographic microscopy is one of the most prominent ones with the possibility of obtaining quantitative phase information by measuring the phase shift change caused by the refractive index of objects. In the lensless digital holographic microscopy system, a pinhole and a light-emitting diode are sufficient to create a holographic pattern on the camera sensor. Here, the optimization of a digital lensless inline holographic microscopy setup was performed to obtain optimal phase value. Also, to retrieve the lost phase information during the recording step, the numerical solution was performed with the single and multi-shot phase retrieval methods. Then, human breast adenocarcinoma (MDA-MB-231) and human myeloid leukemia (U937) cells were analyzed to obtain phase shift, perimeter, and circularity values. These parameters were used to obtain a quantitative differentiation model to replace the traditional labeling or visual confirmation steps with a direct analysis manner. The analysis of respective cells with the classification, object detection, and conditional generative adversarial models can be used directly with pre-trained weights to lessen the computational workloads. With this study, the quantitative analysis with lensless holographic microscopy setup was shown to be a label-free differentiation mechanism to separate cancer cells from monocytes cells which could be used for the early diagnosis of cancer. Also, the proposed method has the potential to be used to identify other cells with links to the diagnosis of different diseases.