Optimization of the automated Sunnybrook Facial Grading System - Improving the reliability of a deep learning network with facial landmarks.

OBJECTIVE: The Sunnybrook Facial Grading System (SFGS) is a well-established grading system to assess the severity and progression of a unilateral facial palsy. The automation of the SFGS makes the SFGS more accessible for researchers, students, clinicians in training, or other untrained co-workers and could be implemented in an eHealth environment. This study investigated the impact on the reliability of the automated SFGS by adding a facial landmark layer in a previously developed convolutional neural network (CNN). METHODS: An existing dataset of 116 patients with a unilateral peripheral facial palsy and 9 healthy subjects performing the SFGS poses was used to train a CNN with a newly added facial landmark layer. A separate model was trained for each of the 13 elements of the SFGS and then used to calculate the SFGS subscores and composite score. The intra-class coefficient of the automated grading system was calculated based on three clinicians experienced in the grading of facial palsy. RESULTS: The inter-rater reliability of the CNN with the additional facial landmarks increased in performance for all composite scores compared to the previous model. The intra-class coefficient for the composite SFGS score increased from 0.87 to 0.91, the resting symmetry subscore increased from 0.45 to 0.62, the symmetry of voluntary movement subscore increased from 0.89 to 0.92, and the synkinesis subscore increased from 0.75 to 0.78. CONCLUSION: The integration of a facial landmark layer into the CNN significantly improved the reliability of the automated SFGS, reaching a performance level comparable to human observers. These results were attained without increasing the dataset underscoring the impact of incorporating facial landmarks into a CNN. These findings indicate that the automated SFGS with facial landmarks is a reliable tool for assessing patients with a unilateral peripheral facial palsy and is applicable in an eHealth environment.

New method for analysing spatial relationships of facial muscles on MRI: a pilot study.

Dysfunction of the facial musculature can have significant physical, social, and psychological consequences. In surgeries such as cleft surgery or craniofacial bimaxillary osteotomies, the perioral facial muscles may be detached or severed, potentially altering their functional vectors and mimicry capabilities. Ensuring correct reconstruction and maintenance of anatomical sites and muscle vectors is crucial in these procedures. However, a standardized method for perioperative assessment of the facial musculature and function is currently lacking. The aim of this study was to develop a workflow to analyse the three-dimensional vectors of the facial musculature using magnetic resonance imaging (MRI) scans. A protocol for localizing the origins and insertions of these muscles was established. The protocol was implemented using the 3DMedX computer program and tested on 7 Tesla MRI scans obtained from 10 healthy volunteers. Inter- and intra-observer variability were assessed to validate the protocol. The absolute intra-observer variability was 2.6 mm (standard deviation 2.0 mm), and absolute inter-observer variability was 2.6 mm (standard deviation 1.5 mm). This study presents a reliable and reproducible method for analysing the spatial relationships and functional significance of the facial muscles. The workflow developed facilitates perioperative assessment of the facial musculature, potentially aiding clinicians in surgical planning and potentially enhancing the outcomes of midface surgery.

The average three-dimensional face for different sex and age groups in a Dutch population.

The increasing use of three-dimensional imaging calls for reference models representing large parts of the population. The aim of this prospective study was to create templates depicting facial maturation in the younger age groups. Healthy Dutch volunteers were captured, without selection of inclusions. Three-dimensional average faces were created using MATLAB, for both genders in four age groups (4-8 years, 8-12 years, 12-16 years, and ≥16 years). Variation within the groups was calculated and depicted on an average face with a green to red colour scale, corresponding to standard deviations between 0 and ≥ 3 mm, respectively. Measurements of the distances of eight peri-oral landmarks were provided as ratios. The statistical analysis was performed using ANOVA and Tukey's test. Three-dimensional reconstructions of the average face and their first principal component were created for each gender and age group. The first principal component comprised the facial width for each group, and the variation of landmarks was low. All ratios showed an increasing trend with increasing age, except for the ratio of philtrum width to mouth width. This study is novel in comparing facial morphology by means of ratios and in creating average faces for the different young age groups. These data provide useful insights into facial maturation, which might be beneficial for facial surgeons.