NeoSCB: Transforming Neonatal Care with AI-Driven Jaundice Screening App for Midwives

A research group formed from the union between the UCL Department of Medical Physics and Biomedical Engineering and UCL Hospitals has geared up to launch NeoSCB.

A research group formed from the union between the UCL Department of Medical Physics and Biomedical Engineering and UCL Hospitals has geared up to launch NeoSCB, significantly enhancing newborn healthcare by developing a smartphone app that screens for severe jaundice through eye photo analysis. The vision of NeoSCB is to equip every midwife in the UK with an accessible jaundice screening tool, potentially saving the NHS over £4.3M annually from unnecessary jaundice blood test referrals. By reducing hospital visits for confirmatory tests, the app also alleviates stress for parents and improves resource allocation within healthcare systems. This innovation not only enhances clinical outcomes but also empowers midwives with a reliable diagnostic tool, ensuring timely interventions that can significantly impact newborn health.

Led by Professor Terence Leung and Consultant Neonatologist Dr. Judith Meek, the NeoSCB team participated in the UCL CDI Impact Accelerator programme. The initiative accelerated their journey from research to practical application by leveraging cloud computing technologies. Despite initial lack of cloud infrastructure experience, the team successfully transitioned their mobile application to AWS, adopting a decoupled serverless architecture.

Adressing Neonatal Jaundice Efficiently

For every 100 babies born, 60 will turn yellow in the first week of life - a common condition called neonatal jaundice, and three will require urgent treatment. Current practice is for midwives to either evaluate babies visually or using expensive equipment, resulting in 200k babies being sent to the hospital unnecessarily for confirmatory blood tests. Avoiding these hospital visits could save the NHS £4.2M a year and parents and babies from the stressful experience. 

The symptoms of jaundice typically present within the first week of a baby’s life. By this time, babies will normally have returned home, and midwives will be responsible for spotting signs of jaundice during home visits. The primary way of identifying jaundice is through visual evaluations by the midwife, but this can be made challenging by factors including babies’ different skin tones and lighting variations. Although some midwives carry specialist equipment with them, these are not commonly available due to the high costs. To ensure a potentially critical illness is not missed, this process often leads to precautionary and unnecessary hospital referrals for blood tests, putting increased strain on the NHS.

The Solution: Empowering Midwives with Advanced Technology

The NeoSCB app enables midwives to screen for significant jaundice during home visits using only a smartphone camera and photos of the baby’s eye, making the diagnostic process accessible and efficient. The NeoSCB team developed a diagnostic algorithm that detects jaundice levels in newborns by capturing eye images and quantifying jaundice using a diagnostic algorithm. Validated through a study of 800 babies in Ghanaian hospitals, the app demonstrated accuracy comparable to commercial transcutaneous bilirubinometers (TcB).

Utilising colour science to measure the yellowness of the sclera (the white of the eye), the neoSCB app estimates bilirubin levels that indicate the degree of jaundice. Babies with levels above 250 micromolar are referred for confirmatory blood tests. This method avoids the challenges of skin pigmentation that can obscure jaundice detection. Midwives can use the neoSCB app on a standard smartphone. After a one-minute calibration with a specialised colour card, the app works in various lighting conditions. During a home visit, the midwife takes pictures of the baby’s eye, and the app’s algorithm identifies the sclera region, estimates bilirubin levels, and provides a referral suggestion within seconds, ensuring timely medical intervention.

Technical Journey with the CDI

Participation in the CDI Impact Accelerator accelerated NeoSCB's research into a fully operational application. Through immersive workshops and expert mentorship, the team navigated cloud computing adeptly, deploying infrastructure using AWS CDK, Amplify, Lambda, Step Functions, and DynamoDB. This enabled them to develop a robust, secure, and scalable mobile app poised to transform neonatal healthcare.

Before the Impact Accelerator, NeoSCB lacked a cloud presence and AWS experience. The 'DevOps – Infrastructure as Code Workshop' provided essential knowledge for constructing their infrastructure with AWS CDK. Using Infrastructure as Code (IaC), they swiftly and securely deployed their application, transitioning from a locally run setup to a decoupled serverless architecture in the AWS cloud.

They subsequently migrated and enhanced their app with AWS Amplify, integrating seamlessly with AWS services like Cognito for secure user management. Adopting a decoupled serverless approach with AWS Lambda, Step Functions, and DynamoDB, they improved automated eye detection algorithms and precise sclera segmentation, enhancing diagnostic capabilities.

After intensive effort in the Impact Accelerator, NeoSCB optimized their machine learning algorithms and built a secure, high-performance mobile app using AWS. Their serverless architecture supports scalability alongside user growth, optimizing costs and expanding accessibility to midwives globally.

Scaling Globally for Enhanced Healthcare Accessibility 

Looking ahead, NeoSCB plans to enhance their algorithms with additional image data, leveraging AWS Rekognition for advanced image processing. These advancements aim to significantly boost diagnostic accuracy and usability, paving the way for widespread adoption and impactful healthcare interventions globally. With plans to make their application globally accessible, this digital solution also has the potential to revolutionise neonatal care in regions with limited healthcare access. This innovation is critical in underserved areas of the global South, where consistent post-hospital midwife contact is often lacking. By addressing undetected health issues such as severe jaundice, NeoSCB's solution promises to have a profound impact, ensuring timely and effective health screenings for all newborns.

The journey of NeoSCB underscores the transformative power of collaborative efforts between academia and industry in healthcare innovation. By equipping midwives with cutting-edge technology, NeoSCB not only enhances clinical outcomes for newborns but also establishes a precedent for scalable and accessible healthcare solutions worldwide.

Co-Founder, Terence Leung, summarises the team's experience with the CDI:

“Before we participated in the CDI Impact Accelerator, we were aware that cloud processing was powerful and could help scaling up digital innovation quickly but we knew very little about how it actually worked and the way to implement a cloud-based app. Learning about the AWS cloud platform was the main reason why we signed up for the scheme. 

Over the past 12 weeks, AWS and ARC cloud architects have introduced our team to a range of AWS services, e.g., Amplify, S3 bucket, Lambda function, Rekognition and SageMaker, guiding us to develop a cloud-based app, helping us to understand the AWS ecosystem. The experience has not only helped this neonatal jaundice screening project but also our other healthcare projects that can potentially benefit from the provision of cloud processing.”