Improving prediction accuracy for neo-adjuvant chemotherapy response in breast cancer through 3D image segmentation and deep learning techniques
K. V. Ranjitha, T. P. Pushphavathi · Artificial Intelligence in Medicine · 2024
Breast cancer is the most prevalent and fastest growing disease in the world. In this paper, neo-adjuvant chemotherapy response is predicted by applying the novel Convolutional Neural Network (CNN) using the I-SPY trial database. Also, an innovative filter can be used in the pre-processing phase by combining various filters like Weiner and Gaussian to remove unwanted noise thereby limiting the high or low frequencies that enhances image boundaries. Extraction of ROI (Region of Interest) in the breast MRI images is a part of segmentation. CNN is used as a part of segmentation along with Content based Image Retrieval, thereby improving the efficiency of the segmentation algorithm. Feature Extraction is done and features like Gray-Level Co-occurrence Matrix (GLCM) and Principal Component Analysis-Scale Invariant Feature Transform (PCA based SIFT) are extracted to determine whether the lesions are cancerous or benign. These features extracted from the segmented images give a better spatial relationship between the image pixels and results in good texture analysis for the tumor images. Significant progress has been made in image segmentation and in conceding tumor lesions from MRI scans. However, segmentation of 3D images remains a frontier, unlike for 2D images. In this review paper, from the I-SPY trial MRI database, we use post-contrast MRI images for segmentation. Various data augmentation techniques are used to prevent the model from over-fitting. It is always feasible to implement the CNN algorithm to predict the chemotherapy responses in patients. The novel CNN algorithm for 3D images used in this paper is useful in clinical management to treat locally advanced breast cancer and yields the best accuracy in predicting patients with pCR(Pathological Complete Response) for Neo-Adjuvant Chemotherapy. Also, the novel CNN algorithm used in the MRI images gives better performance with an accuracy of nearly 98%, thereby classifying the image data better.