Enhanced breast cancer diagnosis using modified InceptionNet-V3: a deep learning approach for ultrasound image...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer (BC) is a leading cause of cancer mortality in women, necessitating accurate and early diagnosis.
• Manual detection of breast cancer is time-consuming and prone to errors, driving the need for automated diagnostic tools.
• Deep learning shows promise in medical image analysis but requires substantial data and computational resources for training.

Purpose of the Study:

• To investigate and compare the performance of various deep learning models for breast cancer classification using transfer learning.
• To evaluate the efficacy of transfer learning in overcoming data and computational challenges in deep learning model training.
• To propose and assess a novel deep neural network integrating features from modified InceptionV3 for enhanced classification.

Main Methods:

• Employed transfer learning with pre-trained models to mitigate computational costs and data requirements.
• Evaluated multiple deep learning architectures: modified InceptionV3, GoogLeNet, ShuffleNet, AlexNet, VGG-16, and SqueezeNet.
• Developed a hybrid deep neural network incorporating features from modified InceptionV3.

Main Results:

• The modified InceptionV3 model achieved the highest classification accuracy at 99.10%.
• The modified InceptionV3 model demonstrated superior performance with a recall of 98.90%, precision of 99.00%, and F1-score of 98.80%.
• All evaluated transfer learning models outperformed baseline approaches on the breast cancer datasets.

Conclusions:

• Transfer learning is a viable and effective strategy for deep learning-based breast cancer classification.
• The modified InceptionV3 model shows significant potential for improving diagnostic accuracy in breast cancer detection.
• The findings support the use of advanced deep learning techniques for more precise and efficient cancer diagnostics.