A Connecticut Children's specialist described an AI-assisted photo-screening program designed to identify infants at risk of craniosynostosis and speed referral to craniofacial teams.

The nut graf: The presenter said AI pattern recognition applied to standardized photos can approximate expert clinical assessment, shorten time to specialist evaluation and increase the share of patients eligible for less-invasive early surgery.

Marcus (surname not provided), a craniofacial specialist, said the program takes standardized phone photos submitted by families or pediatricians and runs them through an AI algorithm that risk-stratifies patients for expedited clinic review. "We've created these AI algorithms...the models are somewhere north of 95% accurate," he said, and described results showing patients reached specialists about 20 days sooner than before the program's use, with a higher proportion treated with minimally invasive procedures.

The presentation covered data-privacy safeguards and model-bias mitigation steps; Marcus said the program preprocesses and deidentifies images and that the team augmented training data to address under-representation of African American patients. He and others discussed telehealth and app-based deployment to extend screening to regions with few craniofacial specialists.

Ending: Marcus said the program aims to scale to regions with fewer specialists and that pilot results show earlier diagnosis and improved access to less-invasive surgery; he highlighted ongoing work to ensure privacy and to correct dataset imbalances before broader roll-out.