A fully automatic framework for evaluating cosmetic results of breast conserving therapy

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cosmetic outcome is crucial for evaluating breast conserving therapy effectiveness and guiding patient treatment choices.
• Current methods for breast cosmesis evaluation are subjective, time-consuming, and lack consistency due to reliance on physician ratings or semi-automated pipelines.

Purpose of the Study:

• To develop a fully-automatic machine learning algorithm for objective and efficient breast cosmetic evaluation.
• To introduce a novel set of breast cosmesis features and a comprehensive dataset for algorithm training and validation.

Main Methods:

• Utilized state-of-the-art Deep Learning algorithms for automated breast detection and contour annotation.
• Developed a novel set of Breast Cosmesis features.
• Created a new Breast Cosmetic dataset with over 3,000 images and human annotations from three clinical trials.

Main Results:

• The fully-automatic machine learning framework achieved performance comparable to existing state-of-the-art methods.
• The proposed algorithm eliminates the need for human input, ensuring objectivity and consistency.

Conclusions:

• The developed fully-automatic algorithm offers a more objective, low-cost, and scalable solution for breast cosmetic evaluation.
• This advancement can significantly improve the assessment of breast treatment outcomes and aid in patient remedy selection.