Improving Breast Cancer Detection and Diagnosis through Semantic Segmentation Using the Unet3+ Deep Learning...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Accurate breast cancer detection and diagnosis are crucial for effective treatment and improved patient outcomes.
• Current diagnostic methods can be subjective and benefit from objective, automated analysis.
• Segmentation models offer a potential solution for analyzing medical images like breast ultrasounds.

Purpose of the Study:

• To evaluate the performance of semantic segmentation models for breast cancer detection and diagnosis using breast ultrasound images.
• To compare the effectiveness of the Unet3+ architecture against other leading segmentation models.
• To assess the models' ability to identify the Breast Imaging Reporting and Data System (BI-RADS) lexicon.

Main Methods:

• Utilized an advanced semantic segmentation method with a deep convolutional neural network.
• Compared six models: Unet3+, FCN, Unet, SegNet, DeeplabV3+, and pspNet.
• Analyzed images from 309 patients (151 benign, 158 malignant tumors).

Main Results:

• The Unet3+ model achieved optimal performance with an average accuracy of 82.53% and an average Intersection over Union (IU) of 52.57%.
• Weighted IU reached 89.14%, and global accuracy was 90.99%.
• Demonstrated the effectiveness of semantic segmentation in analyzing breast ultrasound images.

Conclusions:

• Semantic segmentation models, especially Unet3+, provide remarkable results for breast cancer detection and diagnosis.
• The proposed method offers a potential pathway to more accurate and objective breast cancer diagnosis.
• Improved diagnostic accuracy can lead to better patient management and outcomes.