Curtin University, Kent St, Bentley, Perth, WA, Australia
Jane A. Lieviant
Cancer & Stem Cell Biology Program, Duke-NUS Medical School, Singapore
Sylvia Kam
Genetics Service, Department of Paediatrics, KK Women’s and Children’s Hospital, Singapore; SingHealth Duke-NUS Genomic Medicine Centre, Singapore
Jiin Ying Lim
Genetics Service, Department of Paediatrics, KK Women’s and Children’s Hospital, Singapore; SingHealth Duke-NUS Genomic Medicine Centre, Singapore
Jasmine Chew-Yin Goh
Genetics Service, Department of Paediatrics, KK Women’s and Children’s Hospital, Singapore; SingHealth Duke-NUS Genomic Medicine Centre, Singapore
Weng Khong Lim
Genetics Service, Department of Paediatrics, KK Women’s and Children’s Hospital, Singapore; SingHealth Duke-NUS Genomic Medicine Centre, Singapore; SingHealth Duke-NUS Institute of Precision Medicine, Singapore; Cancer & Stem Cell Biology Program, Duke-NUS Medical School, Singapore; Laboratory of Genome Variation Analytics, Genome Institute of Singapore, Singapore
Gareth Baynam
Rare Care Centre, Perth Children’s Hospital, Perth, WA, Australia
Tele Tan
Curtin University, Kent St, Bentley, Perth, WA, Australia
Duc-Son Pham
Curtin University, Kent St, Bentley, Perth, WA, Australia
Saumya Shekhar Jamuar
Genetics Service, Department of Paediatrics, KK Women’s and Children’s Hospital, Singapore; SingHealth Duke-NUS Genomic Medicine Centre, Singapore; SingHealth Duke-NUS Institute of Precision Medicine, Singapore; Paediatric Academic Clinical Programme, Duke-NUS Medical School, Singapore; Corresponding author.
Leveraging Artificial Intelligence (AI) within the rare disease diagnostic odyssey can facilitate a decrease in diagnostic times and an increase in diagnostic rates. Among the steps involved in the odyssey, this project focused on utilizing AI to automate the standardized capturing of clinical free text into Human Phenotype Ontology (HPO) codes. This research project was conducted at both the KK Women’s and Children’s Hospital (KKH), Singapore and the Rare Care Centre at Perth Children’s Hospital, Western Australia (WA), via the Curtin New Colombo Plan (NCP) Scholarship. The outcome of the project saw the development of a Streamlit web application that utilized two (2) pre-trained AI models – PhenoTagger and PhenoBERT – with a human-in-the-loop design. A case study conducted with ten (10) de-identified clinical reports demonstrated a reduction in the HPO extraction task time from ten (10) to twenty (20) minutes per report to less than five (5) minutes.
