Machine Learning In Metopic Craniosynostosis: Does Phenotypic Severity Predict Long-term Aesthetic Outcome?
Jessica Daviana Blum, MSc1, Justin Beiriger, BS2, Carrie Morales, MD3, Liana Cheung, MBBS1, Dillan F. Villavisanis, BA1, Wenzheng Tao, PhD4, Ross Whitaker, PhD4, Daniel Y. Cho, MD, PhD1, Jesse A. Taylor, MD1, Scott P. Bartlett, MD1, Jesse A. Goldstein, MD2, Jordan W. Swanson, MD, MS1.
1Children's Hospital of Philadelphia, Philadelphia, PA, USA, 2UPMC Children's Hospital of Pittsburgh, Pittsburgh, PA, USA, 3University of Pennsylvania, Philadelphia, PA, USA, 4School of Computing, University of Utah, Salt Lake City, UT, USA.
PURPOSE: Aesthetic deterioration after surgical treatment of metopic craniosynostosis may manifest as bitemporal hollowing (TH), lateral orbital retrusion (LOR), or frontal bone irregularities (FBI). Patients with these complications have reoperation rates of 18-46%. To date, there have been few long-term studies assessing the effect of pre-operative dysmorphology on long-term aesthetic outcomes in this cohort. This study evaluates the relationship between metopic severity and long-term aesthetic outcomes using interfrontal angle (IFA) and CranioRate, a novel metopic synostosis severity measure.
METHODS: Patients with metopic craniosynostosis who underwent bi-frontal orbital advancement and remodeling (BFOAR) between October 2000 and December 2014 were reviewed. Five-year aesthetic outcomes were assessed by attending craniofacial surgeons using blinded three-rater aesthetic grading of clinical photos (n=25). Graders assessed Whitaker score as well as the presence of TH, LOR, FBI, or a "catch-all" category of visible irregularities. Two aesthetic scores were generated: one totaling the number of irregularities present out of 4 (AS) and a second that additionally included the median Whitaker score (ASw) with a maximum of 8. Preoperative CT heads were analyzed using CranioRate, a machine learning algorithm trained to recognize morphologic features of metopic synostosis and generate quantitative severity ratings including metopic severity score (MSS) and cranial morphology deviation (CMD).
RESULTS: Patients underwent BFOAR at mean 9.5 months of age. Preoperative scan was performed at mean 7.6 months, yielding an average MSS of 6.6/10 and CMD of 201/300. Pearson correlation revealed a significant association between all CranioRate parameters and IFA (r = -.58 to -.69, p < .001). Average aesthetic assessment was at 5.5 years (range 4.1 - 7.8) postoperatively. Average AS was 2.7/4.0, average ASw was 5.0/8.0, visible irregularities were noted in 25 (100%), FBI were noted in 16 (64%), TH in 21 (84%), and LOR in 5 (20%) patients. CMD was associated with FBI (r = .454, p = .045) but not TH or LOR. The relationship between AS and MSS trended towards significance (r = .379, p =.06), and MSS was significantly associated with the size of frontal bandeau interpositional graft used (r =.508, p = .026), suggesting larger grafts were employed for more severe deformities. Fisher’s exact test showed a significant association between a milder MSS (≤4) and a lower (≤4) ASw (p =.019) and AS (p=.0198).
CONCLUSION: More severe cases of metopic craniosynostosis show increased rates of long-term frontal bone irregularity and cumulative aesthetic dysmorphologies, but not necessarily temporal hollowing or lateral orbital retrusion. Larger bone grafts were employed in more severe cases, attesting to greater surgical expansion in more severe cases. Stratification of metopic severity showed an association between mild phenotype and favorable aesthetic outcomes.
Back to 2022 Abstracts