This work focuses on developing a system that utilizes the concept of Deep Learning and Convolutional Neural Networks to examine stained tissue samples in hematopathology images. The goal is to enhance the analysis of these images using advanced techniques in machine learning. The goal is to classify these tissue sections into specific lymphoma diagnostic categories. Identifying lymphocytes accurately is essential for diagnosing and treating leukemia, particularly in cases of accurate lymphoblastic leukemia (ALL). To improve the classification of Lymphocytes in microscopic images, the study suggests using Transfer learning of advanced CNN architectures including Conv-Net, ResNet50, MobileNetV2, and VGG19. Our models were initially trained on the dataset from ImageNet and later adapted to work with a customized set of leukemia cell images, with a focus on detecting and classifying lymphocytes. In this approach a dense layer is added to each CNN Model for the final classification. The training process relied on the optimizer named Adamax and used Categorical Cross Entropy to calculate the loss. Techniques like early stopping and model checkpointing were applied. This prevents the models from overfitting and ensures the best performance. The results indicate that the ResNet50 model outperforms the others in classification accuracy and robustness. This makes it a more reliable tool for the automated identification of lymphocytes in leukemia diagnosis. This study supports the use of transfer learning models in clinical diagnostics for improved patient outcomes by highlighting how they compare in terms of increasing the accuracy of medical image classification.