An Automatic Framework for Nasal Esthetic Assessment by ResNet Convolutional Neural Network

Area of Science:

• Medical Imaging
• Computer Vision
• Plastic Surgery

Background:

• Nasal base aesthetics presents a complex challenge in reconstructive and cosmetic surgery.
• Accurate assessment of nasal base symmetry is crucial for successful rhinoplasty and reconstruction outcomes.

Purpose of the Study:

• To develop a novel automatic framework (AF) for evaluating nasal base aesthetics and symmetry.
• To provide a tool that assists in preoperative planning, intraoperative decision-making, and postoperative assessment in rhinoplasty.

Main Methods:

• A hybrid model for nasal base landmark recognition.
• A combined deep convolutional neural network (CNN) and multi-layer perceptron neural network (MLP) model for symmetry prediction.
• Utilized data augmentation techniques to enhance model training on nasal base images.

Main Results:

• The automatic framework (AF) demonstrated results closely correlated with otolaryngologists' ratings.
• The AF effectively predicts nasal base symmetry and identifies asymmetry areas.
• Visualizations confirmed the framework's capability in semantic prediction of nasal aesthetics.

Conclusions:

• The proposed automatic framework offers a valuable tool for objective nasal base aesthetic evaluation.
• The AF can significantly improve the precision and consistency of symmetry assessment in rhinoplasty and reconstructive procedures.
• This deep learning approach facilitates better surgical planning and outcomes in nasal surgeries.