Related Experiment Video
Updated: Sep 20, 2025
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
Improving Breast Cancer Diagnosis in Ultrasound Images Using Deep Learning with Feature Fusion and Attention
Sohaib Asif1, Yuqi Yan2, Bojian Feng1
1Taizhou Key Laboratory of Minimally Invasive Interventional Therapy & Artificial Intelligence, Taizhou Campus of Zhejiang Cancer Hospital (Taizhou Cancer Hospital), Taizhou, Zhejiang 317502, China (S.A., Y.Y., B.F., L.S., D.X.); Center of Intelligent Diagnosis and Therapy (Taizhou), Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences, Taizhou, Zhejiang 317502, China (S.A., Y.Y., B.F., L.S., V.Y.W., D.X.); Wenling Institute of Big Data and Artificial Intelligence in Medicine, Taizhou, Zhejiang 317502, China (S.A., Y.Y., B.F., L.S., V.Y.W., D.X.); Department of Diagnostic Ultrasound Imaging & Interventional Therapy, Zhejiang Cancer Hospital, Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences, Hangzhou, Zhejiang 310022, China (S.A., Y.Y., B.F., T.J., J.Y., L.L., M.S., M.S., L.S., V.Y.W., D.X.).
This study introduces a deep learning model for classifying benign and malignant lesions in ultrasound images, achieving superior accuracy. The model outperforms traditional methods and radiologists, offering a reliable tool for medical diagnostics.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oncology
Background:
- Early detection of malignant lesions in ultrasound images is critical for effective cancer diagnosis and treatment.
- Deep learning models offer potential to enhance accuracy, reduce errors, and improve efficiency in medical image analysis compared to traditional methods.
- This study focuses on the application and interpretability of a deep learning model for classifying benign and malignant lesions in ultrasound images.
Purpose of the Study:
- To develop and evaluate a feature fusion-based deep learning model for classifying benign and malignant lesions in ultrasound images.
- To assess the model's performance against traditional deep learning models and human radiologists.
- To enhance the interpretability of the deep learning model's decision-making process.
Main Methods:
- A feature fusion-based deep learning model was proposed, integrating MobileNetV2 and DenseNet121 architectures with feature fusion and attention mechanisms.
- The model was trained and validated on a private clinical dataset (2171 images) and a public BUSI dataset (780 images).
- Interpretability was achieved using Grad-CAM, Saliency Maps, and shapley additive explanations (SHAP) techniques, with comparative analysis against radiologists.
Main Results:
- The proposed deep learning model achieved high performance, with an AUC of 0.9320 on the private dataset and 0.9834 on the public dataset.
- The model significantly outperformed traditional deep convolutional neural network models and demonstrated diagnostic performance exceeding that of radiologists.
- Interpretability techniques provided insights into the model's focus on relevant image features, supporting its accurate classification.
Conclusions:
- The developed deep learning model demonstrates superior accuracy in classifying benign and malignant lesions in ultrasound images.
- The model's performance surpasses both traditional methods and human expert diagnosis, highlighting its potential as a reliable diagnostic tool.
- The integration of interpretability techniques enhances trust and understanding of the model's predictions in medical diagnostics.
