Predictive Modeling of Neoadjuvant Breast Cancer Treatment Responses: A Comparative Analysis of CNN, Vision Transformer and Machine Learning Models
摘要
A crucial aspect of breast cancer follow-up involves predicting the pathology response to neoadjuvant treatment in female patients. The aim of this study is to assess the Convolutional Neural Network (CNN) and Vision Transformer (ViT) through a thorough analysis of the ACRIN 6698/ISPY2 dataset, which includes images from 292 patients, for predicting the pathological response to neoadjuvant therapies. In the case, machine learning (ML) models are used to assess the importance of feature engineering, specifically feature selection (FS), on the ISPY1 clinical trial dataset. In summary, the Vision Transformer model outperformed the CNN model in this investigation. The second case highlighted the effectiveness of feature selection in our approach, enhancing the performance of ML models.