Facial Plastic and Reconstructive Surgery

AI for Diagnostic Support, Pre-op Planning, and Post-op Evaluation

By Yasmine Madan

Assessing facial symmetry in seconds. Visualizing surgical outcomes before making an incision. Generating age reduction estimates after a procedure. In facial plastic and reconstructive surgery (FPRS), these are no longer fiction – they’re driven by artificial intelligence (AI). AI tools are being integrated in FPRS across various stages, including diagnosis, surgical planning, and outcome assessment (1).

TLDR:

Diagnostic support – Tools such as Emotrics and other computer vision models offer automated, objective assessment of facial symmetry and ear anomalies using 2D images
Pre-op planning – AI tools assist in wrinkle detection, simulate rhinoplasty outcomes, and predict free flap complication risk to guide surgical planning and patient counseling
Post-op evaluation – Software such as FaceReader and other models quantify changes in emotional expression, age, and attractiveness post-surgery, offering objective assessment of aesthetic outcomes

AI Applications in Diagnostic Support

Facial Paralysis Evaluation
Grading systems for facial paralysis, such as the House-Brackmann Grading System (HBGS), often fail to reflect minute changes between stages. Advancements in computer vision (CV) have introduced automated systems that use 2D images to quantitatively assess facial symmetry and movement. For instance, the ‘Emotrics’ software automatically identifies key facial landmarks on 2D front-facing images and analyzes features such as eyebrow symmetry, palpebral fissure width, lip position during smiling, and overall facial movement by measuring distances between these landmarks (2). By detecting sub-millimeter improvements in eyelid function and symmetry that may not be captured by traditional scales such as the HBGS, Emotrics may provide improved surgical planning and outcome measurement (3). The software is currently available as an open-source tool through GitHub and has been implemented at the Facial Nerve Center at the Massachusetts Eye and Ear Infirmary (MEEI) for clinical outcome tracking (4).

Detection of Auricular Anomaliea
Objective assessment of ear deformity is challenging due to the ear’s complex shape. CV model shave demonstrated near-perfect accuracy in providing objective assessments by automatically identifying ear deformity from 2D images (5). These models hold promise not only for diagnostic support, but also for evaluating treatment outcomes. CV models have also been developed to objectively grade microtia severity from images and identify conditions that are often diagnostically challenging such as craniosynostosis (6,7). These models may provide diagnostic support in remote and underserved areas, thereby enhancing early detection and improving access to specialist care.

AI Applications in Pre-operative Planning

Facial Rejuvenation
CV models have been used to quantitatively detect facial wrinkles with over 85% accuracy, aiding in clinical decision-making for procedures such as filler injections (8). Additionally, AI models have shown promise in predicting 3D post-operative outcomes for facial rejuvenation, while also predicting dermal filler volume (8,9). This allows both surgeons and patients to visualize potential outcomes pre-procedure. For example, ‘VISIA’ is a commercially available tool that captures left, right, and frontal facial images, then uses AI to generate detailed assessments of spots, wrinkles, texture, pores, and other skin features (10,11). It has been implemented across Canada and the US, where physicians report enhanced patient communication, education, and post-treatment outcomes tracking (10). However, models for personalized filler volume prediction and outcome simulation are still in development and are not yet integrated into clinical practice.

Rhinoplasty Design
AI models have been developed to facilitate pre-operative planning in rhinoplasty by providing automated pre-operative 3D simulation of results (12,13). One study demonstrated only a 0.8mm deviation between AI-generated and manually designed surgical plans (12). This has the potential to enhance both surgical planning and patient satisfaction. Commercially available tools for rhinoplasty simulation, such as ‘Crisalix’ and ‘FaceTouchUp’, are used by physicians and patients across Canada and the US (14,15).

Free Flap Risk Stratification
Machine learning (ML) models have been created to identify perioperative risk factors and estimate their likelihood of occurrence. For instance, a ML model has been developed to provide an estimate of experiencing a specific complication after microvascular free tissue transfer based on patient information (16). Although these tools are still being refined in research settings, they show promise for enhancing surgical decision-making and improving pre-operative communication with patients.

AI Applications in Post-operative Evaluation

Browlift and Facelift Outcomes

CV models such as ‘FaceReader’ provide automated assessment of facial expressions (17). These tools have been used in clinical settings to measure emotional expression changes after browlifts, detecting objective increases in perceived happiness and reductions in sadness (18). Additionally, similar models have been used to quantify reductions in apparent age after facelifts (19). These tools provide objective measures of procedure effectiveness beyond patient-reported satisfaction and clinical judgement.

Post-Rhinoplasty Outcomes
Software that utilizes CV has been used in research to objectively assess rhinoplasty results, with outcomes showing increases in AI-rated attractiveness and decreases in perceived age (reduction of 1-3 years on average) post-surgery (20,21). This offers surgeons a quantifiable metric for procedural success, but has not yet been widely adopted in clinical practice.

References

1. Park KW, Diop M, Willens SH, et al. Artificial Intelligence in Facial Plastics and Reconstructive Surgery. Otolaryngol Clin North Am 2024; 57: 843-852. 20240708. DOI: 10.1016/j.otc.2024.05.002.

2. Kim MG, Bae CR, Oh TS, et al. Reliability and Validity of Emotrics in the Assessment of Facial Palsy. J Pers Med 2023; 13 20230713. DOI: 10.3390/jpm13071135.

3. Greene JJ, Tavares J, Guarin DL, et al. Clinician and Automated Assessments of Facial Function Following Eyelid Weight Placement. JAMA Facial Plast Surg 2019; 21: 387-392. DOI: 10.1001/jamafacial.2019.0086.

4. Facial Plastic and Reconstructive Surgery. Massachusetts Eye and Ear., https://masseyeandear.org/otolaryngology-outcomes/facial-plastic.
   Hallac RR, Lee J, Pressler M, et al. Identifying Ear Abnormality from 2D Photographs Using Convolutional Neural Networks. Sci Rep 2019; 9: 18198. 20191203. DOI: 10.1038/s41598-019-54779-7.

5. Wang D, Chen X, Wu Y, et al. Artificial intelligence for assessing the severity of microtia via deep convolutional neural networks. Front Surg 2022; 9: 929110. 20220908. DOI: 10.3389/fsurg.2022.929110.

6. Sabeti M, Boostani R, Taheri B, et al. Image processing and machine learning for diagnosis and screening of craniosynostosis in children. Interdisciplinary Neurosurgery 2024; 36: 101887. DOI: https://doi.org/10.1016/j.inat.2023.101887.

7. Alrabiah A, Alduailij M and Crane M. Computer-based Approach to Detect Wrinkles and Suggest Facial Fillers. International Journal of Advanced Computer Science and Applications 2019; 10. DOI: 10.14569/IJACSA.2019.0100941.

8. Shah S, Bennamoun M and Molton M. Machine Learning Approaches for Prediction of Facial Rejuvenation Using Real and Synthetic Data. IEEE Access 2019; PP: 1-1. DOI: 10.1109/ACCESS.2019.2899379.

9. VISIA. Canfield Scientific., https://www.canfieldsci.com/imaging-systems/visia-complexion-analysis/.

10. Henseler H. Investigation of the precision of the Visia(®) complexion analysis camera system in the assessment of skin surface features. GMS Interdiscip Plast Reconstr Surg DGPW 2022; 11: Doc08. 20221129. DOI: 10.3205/iprs000169.

11. Li R, Shu F, Zhen Y, et al. Artificial Intelligence for Rhinoplasty Design in Asian Patients. Aesthetic Plast Surg 2024; 48: 1557-1564. 20230814. DOI: 10.1007/s00266-023-03534-5.

12. Chinski H, Lerch R, Tournour D, et al. An Artificial Intelligence Tool for Image Simulation in Rhinoplasty. Facial Plast Surg 2022; 38: 201-206. 20210529. DOI: 10.1055/s-0041-1729911.

13. Crisalix. Crisalix S.A., https://www.crisalix.com/en.

14. Facetouchup. Facetouchup., https://www.facetouchup.com.

15. Formeister EJ, Baum R, Knott PD, et al. Machine Learning for Predicting Complications in Head and Neck Microvascular Free Tissue Transfer.

16. The Laryngoscope 2020; 130: E843-E849. DOI: https://doi.org/10.1002/lary.28508.

17. FaceReader. Noldus Information Technology., https://noldus.com/facereader.

18. Boonipat T, Lin J and Bite U. Detection of Baseline Emotion in Brow Lift Patients Using Artificial Intelligence. Aesthetic Plast Surg 2021; 45: 2742-2748. 20210927. DOI: 10.1007/s00266-021-02430-0.

19. Gibstein AR, Chen K, Nakfoor B, et al. Facelift Surgery Turns Back the Clock: Artificial Intelligence and Patient Satisfaction Quantitate Value of Procedure Type and Specific Techniques. Aesthet Surg J 2021; 41: 987-999. DOI: 10.1093/asj/sjaa238.

20. Khetpal S, Peck C, Parsaei Y, et al. Perceived Age and Attractiveness Using Facial Recognition Software in Rhinoplasty Patients: A Proof-of-Concept Study. J Craniofac Surg 2022; 33: 1540-1544. 20220314. DOI: 10.1097/scs.0000000000008625.

21. Dorfman R, Chang I, Saadat S, et al. Making the Subjective Objective: Machine Learningand Rhinoplasty. Aesthet Surg J 2020; 40: 493-498. DOI: 10.1093/asj/sjz259.