Hello my name is Shreya Shenoy and I am a junior at American Heritage School of Boca/Delray. Welcome to my website!

Abstract

Acute lymphoblastic leukemia (ALL) is the most commonly found cancer in children in the United States. It develops from lymphoblasts, a type of white blood cell. Due to its rapid progression and the lack of specific diagnosis tests, early detection of ALL is important. A neural network can be a helpful tool to diagnose and differentiate between benign and malignant cells or tumors. Neural networks are algorithms used for deep learning and finding patterns or relationships based on input data. One type of neural networks, called convolutional neural networks, is utilized in image processing and classification. In this project, the researcher created a convolutional neural network based on the SqueezeNet model by pre-processing and loading images from the C_NMC_2019 dataset, creating convolutional layers with Fire modules, applying a ReLU and cross-entropy loss function, and running the model a total of three times each with fifteen epochs. The researcher hypothesized that this neural network would achieve high accuracy rates similar to those of other models used with this dataset in past studies. The researcher’s hypothesis was supported, with the accuracy rates reaching as high as 0.9637 on the training set and 0.7827 on the test set. The ability of the SqueezeNet model to achieve similarly high accuracy rates as other models indicates that it can be beneficial because it saves time and memory. This project demonstrates that neural networks can be useful in various medical areas and the results can be implemented in diagnosing and detecting ALL.