Browsing by Author "Muhammad Omer Aftab"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Classification of Acute Myeloid Leukemia Using a Hybrid Deep Machine Learning Technique
    (UMT, Lahore, 2022) Muhammad Omer Aftab
    Artificial is one of the most emerging computer science domains, that collaborates with other computing domains like Digital Image Processing, Internet Of Things and Data Sciences is causing wonders and a huge difference in the society. AI is being equipped within traditional systems with the intension to introducing autonomy and adding cogitative abilities to enhance the decision-making capabilities of the systems. Cancer is a deadly disease and is one of the biggest leading causes of death in humans, cancer causes the death of thousands of people every year. Cancer had evolved so much and had various types that affects various types of organs and body parts, the symptoms vary from human to human but some of the major symptoms are quite similar to other relative disease, in most cases cancer is diagnosed in middle or last stages. So, the Diagnosis of cancer at the early stages is very important to save lives and to decrease the mortality rate as well. Moreover, cancer treatments are often risky and is very expensive and takes a lot of time to only even for doctors to understand the disease and its variations and causes. Leukemia is known as blood cancer that also consists of a lot to types, Leukemia effects the white blood cells which are the natural human disease defence mechanism. cancer driver genes mutation prediction is thus helpful to control the death rate due to cancer. Identification of leukemia driver mutation sites through experimental mechanisms can be expensive, slow and laborious. The proposed methodology presents a state-of-the art hybrid technique using 2 modules for the detection and classification of AML, both of the modules are connected via decision-making where the first module uses binary classification method for the classification of AML through Gene expression of genomic sequences while the second modules uses a multi-class classification technique that classifies AML into 15 classes. If the patient is predicted to have AML via his/her genomic sequence, then the system will move to the next module that will take the leukemia images as an input and process them and classify the images into the respective classes. Both of the modules have produced really good and satisfactory results. The proposed methodology can be proven to be very useful in the domain of bioinformatics for the detection and classification of Leukemia in early stages. The methodology produces a training and validation accuracy for the AML gene classification is 100% and 99% respectively. Moreover, for the AML image classification the model produces 100% and 96.7% of training and validation accuracy respectively