Blood disorders, including leukemia and infectious diseases of red blood cells, pose diagnostic challenges due to their complex presentations and reliance on expert-driven cytomorphological analysis. While DL has shown promise in automating medical imaging, its dependence on large annotated datasets and limited interpretability hinder clinical integration. This study introduces a novel framework that integrates topological features with DL techniques to enhance cytomorphological image analysis. By capturing global topological structures alongside localized DL patterns, our approach improves diagnostic accuracy, robustness, and interpretability, particularly in limited data settings. Experimental results demonstrate superior performance, addressing key limitations and advancing reliable, efficient blood disorder diagnosis.