Deep Learning-Based Multimodal Breast Cancer Detection
Nithya Shankar S, R Adithya, Iyappan Ramalakshmi Oviya, S. Vaishnavi, Veena Sundari A, Anagha Rajan · 2025
Breast cancer identification is important for the early diagnosis. Traditional diagnostic methods such as mammography, ultrasound, and thermography often have problems like false positives, operator dependence, and low sensitivity in dense breast tissue. While recent deep learning advancements have shown great promise in medical imaging, most studies have focused on using one model, which limits diagnosis accuracy. To improve breast cancer categorization, we investigate a variety deep learning architecture that includes mammography, ultrasound, and thermography. This paper explores convolutional neural networks (CNNs) such as ResNet50, DenseNet121, EfficientNetB0, and InceptionV3 models for prediction of cancer. Results tell that DenseNet121 showed the highest accuracy for the ultrasound dataset 83.86%, InceptionV3 showed highest accuracy for Thermography and Mamography dataset with 79.60% accuracy and 83.58%.