Haematological disorders that can be diagnosed by blood cell testing provide problems for biomedicine. AI might automate health monitoring and diagnosis, which is useful because professional laboratory diagnoses are expensive and time-consuming. Many methods have been developed for deep learning (DL) and machine learning (ML). Due to its accuracy, automation, and low error rates, DL-based research methods are growing in popularity. However, the current DL methodology must address computational inefficiencies, intra- and inter-class variabilities, and other issues. A unique method for automatic blood cell classification to anticipate haematological disorders uses microscopic images. The program uses DL to classify haematological diseases more accurately. The proposed DL model classifies haematological disorders using raw microscopic blood cell pictures. The proposed model uses EfficientNet, feature engineering, and microscopic image preprocessing. The innovative preprocessing method reduces microscopic picture diversity in both classes. A pre-configured DL model’s expanded layer design simplifies autonomous feature engineering and categorization, reducing computational resources. We compare the proposed DL model to industry standards using a blood cell dataset. Compared to existing approaches, the proposed model reduces processing time by 20.14% and increased blood cell classification accuracy by 2.93%.