Acute Lymphoblastic Leukemia (ALL) is a rapid form of cancer that can be fatal if not detected and treated as quickly and accurately as possible. Microscopy based manual diagnosis is the traditional way of diagnostic process, which may be time consuming and with high variation according to observer. To overcome these drawbacks, we present a deep learning pipeline with an improved YOLOv8 model to detect and classify ALL cells from blood smear images using only positive samples. This much quicker and efficient model has the power to be used for faster diagnosis when time in medical field is most crucial. The following method was applied to fine-tune the model (feature extraction): All of these, resulted in 98% accuracy for all cells and sensitivity specificity and precision values around 91% when distinguishing ALL from noncancerous controls. Conclusion: This novel high-performance detection system decreased both diagnostic error and processing time, representing an effective alternative to usual procedure. We aim to automate the detection which may provide a clinical solution scalable towards assisting pathologists and oncologist make more reliable decisions faster. In this setting accurate differentiation of cancerous and non-cancerous cells is crucial for clinical event classification leading to better patient outcome, especially in resource-limited settings with limit access to expert assessment. Summary: Our newly developed deep learning model for hematological diagnostics is a great improvement in leukemia detection, which can contribute to the wider mixing of AI into healthcare through its improved efficiency and automation. ALL ARE CORDIALLY INVITED