Investigating blood cell images for enhanced hematologic disorder detection using multi-scale feature learning with a hybrid deep learning model.

Haematological diseases, also known as blood disorders, encompass a range of conditions that affect the blood and its components, including platelets, red blood cells, plasma, white blood cells, and the bone marrow. These illnesses can disrupt the normal function, production, or quantity of blood cells, leading to severe health problems. Standard types of haematological diseases include lymphoma, anaemia, multiple myeloma, leukaemia, thrombocytopenia, and haemophilia. Diagnosis typically involves blood tests and microscopic examinations of bone marrow or blood samples. At the same time, treatment varies according to the form and severity of the disease, often including chemotherapy, medication, or bone marrow transplantation. Screening for haematological disorders is crucial for their diagnosis and effective treatment. The medical diagnosis of a haematological disorder is mainly dependent on laboratory blood tests. Still, the most experienced haematology professional can supervise designs, relations, and deviations among the growing number of blood parameters that modern laboratories compute. Deep learning has experienced significant growth over the last decade and is being applied effectively in various intelligent applications, including the detection of haematological disorders. This paper presents a Multi-Scale Feature Learning-Based Enhanced Hematologic Disorder Detection Using Hybrid Deep Classification Model (MSFLHDD-HDCM) approach. The goal is to develop an accurate model for detecting and classifying hematologic disorders using microscopic images of blood cells. Initially, the image pre-processing stage employs the geometric mean filter to enhance image quality by effectively removing noise and standardizing the input data, thereby improving the accuracy of subsequent analysis. Furthermore, the MSFLHDD-HDCM method utilizes the inception modules technique for feature extraction. Finally, a hybrid model combining a convolutional neural network and a bidirectional gated recurrent unit is utilized for classification. A comprehensive experiment of the MSFLHDD-HDCM method is performed under the blood cell image dataset. The comparison analysis of the MSFLHDD-HDCM method demonstrated a superior accuracy value of 99.67% over existing techniques.