Deep Learning-Based Multiple Myeloma Identification Using Micro-imaging Technique
Shamama Anwar · 2023
Multiple myeloma (MM) is a category of cancer in which the plasma cells are affected. The medication prescription for MM depends on its stage, i.e., the spread of the disease in the body. Early diagnosis is vital for effective treatment, recovery, and reducing mortality. Advancement in computing technologies has led to the advent of computer-aided diagnosis. Further employing deep learning methodologies and tools in medical image analysis has given profound results that lead to quick and accurate analysis. This chapter utilizes the deep learning technique to design an automated diagnostic system to detect MM using micro-imaging, i.e., from microscopic blood-stained images. Since deep learning techniques require ample data for efficient training, it is a prerequisite to obtain these data, which is also a mandate to minimize the occurrence of overtraining. For this reason, diverse data augmentation techniques are employed to enlarge the dataset size. The model is trained on a total of 574 images (after augmentation) and tested on 246 images, achieving 92.8% accuracy averaged over a series of ten trials. Prior pre-processing or segmentation is not required, and the model is capable of working on raw data efficiently. The proposed deep learning architecture can thus be beneficial for domain users in detecting MM efficiently.