Development and Validation of a Convolutional Neural Network Model for Early Detection of Invasive Ductal Carcinoma in Histopathological Images (Preprint)

Abstract

Breast cancer is a major health concern for women worldwide and poses a significant challenge to healthcare systems. According to the World Health Organization(WHO), 2.3 million women were diagnosed with breast cancer in 2020, resulting in 685,000 deaths. Invasive Ductal Carcinoma (IDC) accounts for 80% of these cases, making it crucial to accurately diagnose it in a timely manner. Traditional breast cancer detection methods rely on diagnostic imaging techniques like mammography, ultrasound, and MRI, which are interpreted by trained radiologists. However, these methods’ accuracy depends on the radiologist’s experience and can be subjective, leading to variability in diagnosis. False positives and negatives are not uncommon, which can result in missed cancers or unnecessary biopsies. This paper proposes a Deep Convolutional Neural Network (DCNN) based model to detect IDC in histopathological images of breast tissue. The model uses a robust dataset and fine-tuned hyperparameters, highlighting the potential of deep learning in improving diagnostic accuracy in oncology. The model achieved an 87% accuracy on the test set.