Development of an automatic scoring software tool utilising artificial intelligence to measure a cleft appearance outcome for children born with a complete unilateral cleft lip

Lead researcher: Bruce Richard

Institutions: Birmingham Children's Hospital, Birmingham Women's and Children's NHS Foundation Trust and Edge Hill University, Ormskirk

Repair of a cleft of the lip leaves a scar, and nearly always some visible difference in the appearance of the lip such as asymmetry. Such differences in appearance can cause significant distress so it is important that surgery results in the ‘best’ outcome possible.

Parents of children born with a cleft lip want to have confidence that their cleft team will do the best technical operation according to the best protocol to get the best outcome in terms of facial appearance.

However, in order to know which method of lip repair gives the best outcomes, it is necessary to have a way of measuring how good the outcome of a lip repair is. We know how to measure other important outcomes in cleft, such as psychological, speech, dental and facial growth outcomes, in a robust scientific way, yet we are unable to do the same for facial appearance. At the moment studies which look at how good the outcomes of cleft lip repair are mainly use photographs. They measure the quality of the outcome by asking viewers to rate the quality of the repair from the photograph. This is a subjective way of measuring and it gives unreliable results which are not useful for clinical use.

In the last 12 months the team of computer scientists working on this project have succeeded for the first time to create an automatic facial appearance outcome score that uses Artificial Intelligence (AI) to score facial photographs of children with a cleft lip five years after they had the operation. It uses the biases of the subjective human scores to create a far superior outcome by transfer learning in a vision convoluted neural mathematical network of artificial intelligence. The results were presented at the Craniofacial Society of Great Britain and Ireland in April 2024 and now a post-doctoral researcher will refine the AI with new data sets donated to the project by other international centres. The aim is to develop an easy-to-use automated software tool that can score how a person’s face looks after cleft lip surgery. If successful, this will help cleft teams assess which technical method, timing, or surgeon gets the best result.

Project Update:

In the last 12 months the team of computer scientists working on this project have succeeded for the first time to create an automatic facial appearance outcome score that uses Artificial Intelligence (AI) to score facial photographs of children with a cleft lip five years after they had the operation. It uses the biases of the subjective human scores to create a far superior outcome by transfer learning in a vision convoluted neural mathematical network of artificial intelligence. 

Currently post-doctoral researcher, Dr Jonathan Lumentut, is developing the next level of AI using new data sets donated to the project by other international centres. He has completed a data clean of the original 238 images and is working on new algorithms and neural networks to create a better automatic scoring system. He makes week on week improvements to the algorithm. The aim is to develop an easy-to-use automated software tool that can score how a person’s face looks after cleft lip surgery. If successful, this will help cleft teams assess which technical method, timing, or surgeon gets the best result.

Dr Ambika Chada has joined the project. She is drafting a clinical paper for publication, and has submitted an abstract for the 15th International Cleft Congress in Japan, October 2025.

We are exploring new research grant applications to take this project to a clinical application stage.