International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 12 Issue: 09 | Sep 2025
www.irjet.net
p-ISSN: 2395-0072
Transfer Learning-Enhanced Ensemble for Accurate Breast Cancer Diagnosis from Mammography Anurag Kumar Patel1, Rajkumar Sharma2, Vivek Richhariya3 1Student, Department of Computer Science & Engineering, LNCT Bhopal 2,3 Professor, Department of Computer Science & Engineering, LNCT Bhopal
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Breast cancer represents a major worldwide
[3]. Mammography serves as the corner- stone and most
health challenge, ranking among the primary causes of mortality in the female population Achieving precise earlystage detection is essential, as this greatly enhances survival prospects and supports the implementation of more effective treatment approaches. Mammography remains the most prevalent imaging modality for initial screening, though its manual interpretation can be complex and susceptible to inter-observer variability. To address this challenge, this study introduces a deep learning-driven framework for automated breast cancer identification, utilizing mammographic imagery from the MIAS dataset. Three distinct models were conceptualized and assessed: (i) a standard Convolutional Neural Network (CNN), (ii) a hybrid CNN-DNN model enhanced by transfer learning, and (iii) an ensemble CNN-ML model. The latter, which proved most effective, specifically incorporated feature extraction from pre-trained networks, including InceptionV3, ResNet50, and VGG16, followed by a voting-based classification executed with Support Vector Machine (SVM) and Random Forest (RF) algorithms. A thorough assessment of the proposed models revealed that the ensemble CNN-ML framework delivered the strongest classification performance. Notably, this leading model achieved an impressive accuracy of 98.48% in the image-based breast cancer detection task.
Fig -1: Number of cancer cases among women in India widely adopted imaging modality for routine breast cancer screening and detection [4]. This low-dose X- ray technique excels in visualizing internal breast structures and identifying abnormalities such as microcalcifications and masses, making it the primary and most cost-effective tool for early diagnosis and a key contributor to reducing mortality rates [5]. Despite its critical role, the manual interpretation of mammographic images presents considerable challenges due to the complex nature of breast tissue, varied tumour appearances, and the inherent subjectivity and inter-observer variability among radiologists [6]. These challenges may lead to false positives—causing unnecessary patient distress and additional procedures—or false negatives, which can postpone essential treatment [7]. To overcome these hurdles, Computer-Aided Diagnosis (CAD) systems have emerged as vital tools, enhancing diagnostic accuracy and efficiency. Recent progress in Artificial Intelligence (AI), particularly in deep learning (DL), have revolutionized medical imaging diagnostics by automating feature extraction and pattern recognition from raw images, often achieving performance comparable to or exceeding human experts. Deep learning models offer promising solutions for more precise, reliable, and consistent breast cancer detection, leveraging their capacity to discern subtle anomalies that might be missed by manual inspection [8]. Our primary contributions are:
Key Words: Breast Cancer Diagnostics, Medical Imag e Analysis, Deep Learning for Medical Imaging, Transf er Learning Applications, Ensemble Learning Framew orks
1. INTRODUCTION Breast cancer (BC) represents a critical global health issue, acknowledged as the most widespread cancer among women worldwide and a major cause of mortality [1]. Projections indicate a continuous rise, with the World Health Organization estimating 19.3 million cancer cases by 2025 [2]. Figure 1 illustrates the number of cancer cases among women in India. Crucially, early and precise detection is the most significant factor in improving patient survival rates and enabling effective treatment outcomes
© 2025, IRJET
|
Impact Factor value: 8.315
|
ISO 9001:2008 Certified Journal
|
Page 60