A digital space within the hospital to help build artificial intelligence

The team at GOSH DRIVE have been working to implement a new digital space for building and testing artificial intelligence (AI) tools. This vital infrastructure marks a significant step towards the future of digital innovation within GOSH.

Unveiling the Power of Big Data

The new space enables the processing of vastly larger volumes of data, which is crucial for two key purposes:

• Training AI Tools: All AI tools, regardless of their specific function, need to learn which features from the data are important. Data scientists embed these instructions within the AI's code, enabling it to identify patterns and make predictions.
• Rigorous AI Testing: We evaluate AI tools to ensure they are working accurately. Depending on the intended use case, extensive further testing may be necessary to guarantee optimal performance before deployment.

 Securing Sensitive Healthcare Data

Building and testing AI tools in the healthcare domain presents unique challenges due to the sensitive and personal nature of data. Our new digital space addresses these concerns by providing separate, secure environments for identifiable and non-identifiable data.

Development and testing of AI tools occur within the non-identifiable data environment. The tools can then progress to the secure environment containing identifiable data for further testing under strict access controls. These measures ensure that all patient information remains completely confidential and protected. Additionally, only authorised GOSH personnel hold access to the identifiable data environment.

The new digital space in action at GOSH

We’re currently focusing on natural language processing (NLP), an AI technique that transforms unstructured text, like comments in PDFs, into structured, usable data.

Currently, valuable information lies buried within unstructured data, making it difficult to utilise. NLP's data extraction capabilities enable us to make the data readily available for research and innovation. This valuable data can help with discovering new treatments, personalising patient care, and ultimately enhancing patient experience.

Early Impact

The digital space has already significantly enhanced our capabilities, demonstrated by the following examples:

• Streamlined NLP Tool Development: By adopting best practices in software engineering and Reproducible Analytical Pipelines (RAP), our teams are building reusable, efficient, and high-quality NLP tools.
• Drastically Reduced Processing Times: For example, through continuous optimisation the time taken to process 36,000 documents has been reduced from 4 days, to around 10 hours.
• Enhanced Developer Efficiency: Developers can now work on multiple NLP tools simultaneously, fostering the creation of more robust and scalable pipelines.

Looking Ahead

We are committed to continuously improving the digital space by adding new tools and functionalities. This marks the first step in building a future-proof digital infrastructure at GOSH, laying the foundation for ongoing advancements in healthcare through AI.

Working in collaboration

GOSH has a long history of working in collaboration with industry partners to help develop better treatments and therapies, faster. As part of our 5 year collaboration with Roche we have been exploring how NHS and Industry can work together to enhance our data capabilities and infrastructure and improve outcomes for patients. Roche does not have access to any personal information and no data leaves the GOSH DRIVE infrastructure.

This project has been completed by Great Ormond Street Hospital NHS Foundation Trust and Roche Products Ltd as part of a collaborative working agreement. Roche Products Ltd had no influence on the results or decision to publish regarding this work. M-GB-00016519 | March 2024