Classification of Thalassemia Patients Using a Fusion of Deep Image and Clinical Features

It is difficult to diagnose different types of diseases of the same category using only clinical reports, which is why doctors diagnose this type of disease using clinical reports as well as blood smear image of patients. This paper, therefore, come up with a technique that uses both clinical report and blood smear image of patients for detection of thalassemia. Clinical features are extracted by the blood analyzer, and image features are extracted from the blood smear image via the Deep Confusional Neural Network (CNN) and then both features are combined to make a relevant feature set. But this set of features has some redundant features that increase computational complexity. This work uses principal component analysis (PCA) to overcome redundancy in features, resulting in lower computational complexity. This paper used machine learning classification algorithms such as Naive Bayes, random forest (RF) and KNN to classify patients into a thalassemic and normal person with the help of combined features and got accuracy 99 ±1%, specificity 100%, and sensitivity 100%.