Comparative Analysis of Custom vs. Pre Trained CNN Models for Breast Tumor Classification

Read full text Summarise with VIRAT
Authors: Mohammad Aman Ullah Khan, Irfan Sadiq Rahat
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Breast tumor classification is crucial for accurate diagnosis and effective treatment planning in breast cancer patients. Traditional diagnostic methods, although informative, are labor-intensive and subject to variability. This study proposes a novel deep learning (DL) framework to enhance breast tumor classification using histopathological images from the BreakHis dataset. The approach involves a comprehensive preprocessing pipeline, including image resizing, data partitioning, and augmentation techniques to improve model robustness. A custom CNN is designed for feature extraction and classification, outperforming pre-trained models such as MobileNetV1, EfficientNetB1, VGG19, and ResNet50V2. The custom CNN achieves a high accuracy of 93.12% and a ROC-AUC of 0.9708. Hyperparameter optimization is performed using Hyperband and Bayesian Optimization, further fine-tuning the model's performance. The Adam optimizer facilitates model convergence during training. Results indicate that the proposed model significantly challenges or exceeds recent state-of-the-art approaches, offering a valuable tool for breast tumor diagnosis and contributing to improved clinical decision-making.

Fields of Study & Subjects

AI in cancer detection Digital Imaging for Blood Diseases Artificial Intelligence in Healthcare

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
26 April 2026
Language
EN