A Deep Analysis of Transfer Learning Based Breast Cancer Detection Using Histopathology Images
Problem
Breast cancer, common and dangerous in women and men, needs accurate detection from histopathology images containing phenotypic data.
Why it matters
Early, accurate breast cancer detection directly affects patient treatment and survival.
What the AI did
Pre-trained transfer learning models (ResNet50, ResNet101, VGG16, VGG19) classified 2453 histopathology images as invasive ductal carcinoma or non-IDC.
Results
ResNet50: 90.2% accuracy, 90.0% AUC, 94.7% recall, 3.5% loss
Impact
ResNet50 outperformed other tested models on this dataset; no clinical deployment stated.
Record
| Venue | International Conference on Signal Processing and Integrated Networks (SPIN) |
|---|---|
| Published | 2023 |
| Work type | conference-paper |
| Authors | Md Ishtyaq Mahmud, Muntasir Mamun, Ahmed Abdelgawad |
| Citations | 44 (OpenAlex, at import time) |
| References | 29 |
| Model or method | ResNet50 (vs ResNet101, VGG16, VGG19) |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1109/spin57001.2023.10117110 |
| Open access | green |
| Keywords | breast cancer, histopathology, transfer learning, ResNet50, deep learning, IDC |
Importance rating
major — millions of lives, an industry or a field
Abstract states breast cancer is one of the most common and dangerous cancers in women and can afflict men; early accurate detection affects treatment and survival for large populations.