An LSTM Based CNN Model for Multiple Myeloma Disease Detection

Multiple myeloma is characterized by the abnormal proliferation of plasma cells in the bone marrow. A common diagnostic approach involves bone marrow aspiration, where the obtained slides are either visually examined or analyzed using digital image processing software to identify myeloma cells. This study investigates the effectiveness of a CNN-LSTM hybrid model for detecting multiple myeloma. MiMM_SBILab Dataset consisting of 90 images are used for the evaluation of the model. The dataset was manually annotated using Instance Segmentation Image Annotation software. This technique helps in identifying individual myeloma cells and differentiating them from overlapping or adjacent cells. The deep learning model was trained in monitoring training and validation loss for each epoch, with the best model chosen based on minimal validation loss. Results indicate that the CNN-LSTM hybrid model effectively addresses many challenges associated with multiple myeloma segmentation.