K-nearest neighbors algorithm (KNN) and artificial neural networks (ANN) accurately predicting malignancy of breast cancer (BC)
Date
2020
Authors
O'Shea, Bailey
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
With the reoccurrence of unnecessary open surgeries on potential malignant tissue, there is a need for additional non-invasive tools oncologists and radiologists can utilize to help argue the reason behind performing surgical biopsies. Thus, machine learning algorithms (MLAs) have seen a great deal of attention to the classification of tissue malignancy. One major benefit rises in having the ability to utilize past accessible datasets to accurately predict/classify new data/patients with similar features. The purpose of this paper was to apply and assess two MLAs—k-nearest neighbor (KNN) and artificial neural network (ANN)—on classification accuracy of breast cancer (BC) malignancy. Importantly, features used for the MLAs are acquired from imaging modalities, solely. For this particular dataset, features seen to be extracted from medical images include clump thickness, uniformity of cell size, uniformity of cell shape and marginal adhesion. The optimal k-nearest neighbor and ANN hidden layer will be reported. After implementing and testing the two MLAs, the accuracy for the KNN and ANN were 100% at 132-nearest neighbors and 95.24% ± 0.224 respectively. Considering the performance across both MLAs, the optimal classification algorithm for this dataset is the KNN algorithm. Thus, allowing for the possibility of clinical use as an additional consultation tool.
Description
Keywords
Machine Learning Algorithms, Cancer, Oncology, Artificial Neural Networks
Citation
Bailey, O'Shea. K-nearest neighbors algorithm (KNN) and artificial neural networks (ANN) accurately predicting malignancy of breast cancer (BC). 2020.