Predicting Invasive Ductal Carcinoma Based on Convolutional Neural Network

• Published in: 2022 3rd International Conference on Electronic Communication and Artificial Intelligence (IWECAI) pp. 388 - 391
• Main Author: Yu, Junhan
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.01.2022
• Subjects: Breast cancer; component; Computational modeling; Convolutional Neural Network; Convolutional neural networks; Invasive Ductal Carcinoma; Kaggle; Medical services; Pathology; Predictive models; Semantics
• DOI: 10.1109/IWECAI55315.2022.00082

Since the last century, the diagnosis of cancer has been a difficult problem among the world. Currently, the diagnosis of cancer is mainly made by medical experts that they need to speed a lot of time to observe pathological patches. However, this method requires tedious work and is insufficient for diagnosing all patients in low- and middle-income countries that lack medical resources. Which will result the delays in the diagnosis of the patient's condition and decreases the survival rate of patients. Therefore, I created a Convolutional Neural Network (CNN) model to automatically diagnose Invasive Ductal Carcinoma (IDC) in order to help medical experts reduce tedious work. I created a CNN-based model with nine 2D-convolutional layers and three max-pooling layers. The data passing through these layers will eventually output a dichotomous result, i.e., IDC or non-IDC. Then, I chose a dataset of 162 patients from Kaggle to test the performance of the model. The result is satisfactory, the model has 88% accuracy and 95% Area Under Curve (AUC) score in the automatic diagnosis of IDC. Based on this promising result, the model performance well to solve the problem that automatically diagnose IDC for the purpose of assisting medical expert.