Research on an advanced neural network has led to the identification of crucial reference points for breast surgery, opening the potential for an objective assessment of breast symmetry, as reported in a study published in the February issue of Plastic and Reconstructive Surgery®, the official medical journal of the American Society of Plastic Surgeons (ASPS). The journal is published in the Wolters Kluwer Lippincott portfolio.
„Neural networks and machine learning have the potential to improve the assessment of breast symmetry in reconstructive and cosmetic breast surgery and enable quick, automated detection of features used by plastic surgeons.“
Nitzan Kenig, MD, lead author, University Hospital of Albacete, Spain
Development of Neural Networks for Objective Breast Assessment
Breast symmetry is a primary concern in breast surgeries and is generally evaluated through simplistic subjective assessments by both patients and surgeons. Computer programs can enable more objective assessments, although with limitations such as the need for manual data entry and prolonged processing times.
Neural networks, a form of artificial intelligence designed to mimic how the human brain processes data, are being investigated for their potential to improve care in various medical practice areas. Dr. Kenig and colleagues developed an „Ad-hoc Convolution Neural Network“ to detect vital breast features used in the assessment of breast symmetry.
Using an open-source algorithm called YOLOv3 („You Only Look Once“, version 3), the researchers trained their neural network to identify three anatomical features used in assessing the female breast: the breast borders, the nipple-areola complex, and the suprasternal notch (the indentation at the base of the neck, at the top of the breastbone).
The neural network was trained using 200 frontal photos of patients who underwent breast surgery. Its performance in identifying key breast features was then tested using an additional set of 47 photos of patients who underwent breast reconstruction following breast cancer surgery.
Potential for „Quick, Automated, and Objective“ Breast Symmetry Assessment
After training, the neural network displayed a high accuracy rate of 97.7% in localizing the three features. The accuracy for the right and left breast borders and the nipple-areola complex was 100%. For the suprasternal notch, the recognition rate decreased to 87%. Processing was swift, with an average detection time of 0.52 seconds.
The neural network was capable of identifying the key features even in visibly asymmetric breast reconstructions. The high success rate confirmed that the training dataset was sufficient, and no data augmentation techniques were required.
„Neural networks and machine learning have the potential to improve the assessment of breast symmetry in the field of plastic surgery by automatically and swiftly recognizing features used by surgeons in practice,“ conclude Dr. Kenig and his co-authors. They believe that with further advancements in image recognition capabilities and their applications in breast surgery, neural networks could play a role in the assessment of breast symmetry and the planning of both aesthetic and reconstructive plastic surgery.