¹û¶³Ó°Ôº

Phenopolis Ltd, a ¹û¶³Ó°Ôº HealthTech spinout, joined cohort 5 to enhance and expand their groundbreaking digital solution, TrialSense, setting a new standard in the clinical trial landscape.

Phenopolis specialises in building transformative health-tech web applications that comply with regulatory standards, within a secure, distributed, cloud-based environment. Founded in 2018 by Ismail Moghul, a PhD student at the ¹û¶³Ó°Ôº Cancer Institute and Dr Nikolas Pontikos, a postdoc from the ¹û¶³Ó°Ôº Institute of Ophthalmology, Phenopolis recently joined cohort five of the CDI Impact Accelerator to advance and scale up their innovative digital solution, TrialSense. 

TrialSense seamlessly integrates diverse clinical trial data, including imaging, genomics as well as electronic Case Report Forms, with robust data analysis capabilities into a unified system while maintains rigorous auditing standards and data quality controls, enabling users to address data issues early in the trial lifecycle before the data lock. This positions TrialSense to accelerate and revolutionise the clinical trial landscape. 

Currently deployed in over 150 NHS hospitals, TrialSense is having a transformative impact on healthcare research and clinical trial management. During their CDI experience, Phenopolis achieved a significant milestone with the development and deployment of VisionSense, an AI-powered medical image annotation module for TrialSense. VisionSense accelerates the annotation and quantification of biomarkers within images, effectively streamlining processes and reducing time and administrative burdens for pharmaceutical companies, hospitals, and academic institutions involved in clinical trials. This advancement translates into quicker and more accurate identification of treatment safety and effectiveness for trial participants. 

Addressing the Challenge

Every year, over 2,000 new clinical trials are initiated in the UK to determine whether medical products, such as drugs or devices, are safe and effective for patients. However, more than 70% of these trials fail due to data errors. Modern clinical trials collect vast amounts of data, including medical imaging, genomics, and patient-reported data, requiring complex processing. Traditionally, this data is siloed, making integration difficult and error-prone, delaying critical decisions and increasing trial risks. 

The Solution

Phenopolis Ltd has developed TrialSense to enhance the quality of trial data through automation and artificial intelligence, simplifying the ingestion and validation of large amounts of clinical data. By employing real-time data analytics and AI, the platform automatically highlights and provides potential solutions to data issues from multiple sources as soon as they are uploaded. Medical images can be automatically annotated using the client’s segmentation model or interactively annotated using a custom foundation model that continuously improves as further annotations are made. TrialSense also incorporates advanced data analysis capabilities into the platform, bridging the gap between Clinical Trial Management Systems (CTMS) and research, ‘making trials smarter, better, faster, safer’.  

With this innovative solution, clinical trials can integrate a variety of data into a single modular, fully integrated platform in an audited manner. Advanced automation ensures that each data type is processed as required, as soon as it is uploaded. Automated AI-powered quality control reviews all uploaded data, automatically identifying potential issues with data quantity and quality in real time. This enables early and effective remediation, contrasting sharply with current standards where quality control is often performed after the data lock, which is usually too late to rectify any potential issues. This dramatically improves the efficiency of clinical trials, reducing the risk of failure and enhancing overall data integrity. 

Technical Journey with the CDI

During the CDI Impact Accelerator, the Phenopolis team focused on advancing VisionSense, an AI-powered image annotation module within the TrialSense system. VisionSense is designed to accelerate the annotation and quantification of biomarkers in medical images. By leveraging a foundation model, expert graders can interactively annotate images by selecting (and de-selecting) areas of interest with a single click. Through few-shot learning, VisionSense further customises this foundation model as further manual annotations are made, and thereby adapting the underlying foundation model to the specific biomarker of interest. Additionally, VisionSense employs pixel tracking along with the foundation model to propagate annotations through a 3D volume. VisionSense helps enhance the accuracy and efficiency of the annotation process. Key milestones included: 

• Migration to AWS SageMaker Inference: Transitioning existing AI models from AWS Lambda to AWS SageMaker Inference with GPUs to enhance scalability and performance. 
• Enable Interactive Annotation within the TrialSense Annotator: By using Foundation Models, areas of interest can be selected with a single click.  
• Enable 3D volume Annotation Propagation within the TrialSense Annotation: Enhancing annotation capabilities by propagating annotations across 3D volumes by utilizing AI-enabled pixel tracking, improving accuracy and efficiency in biomarker quantification. 
• Collaborative Development of GenomeSense: Collaborating with AWS Solution Architects to establish infrastructure and pipelines for GenomeSense, facilitating comprehensive genomic data processing within the platform. 
• Add Support for Next Flow-based Pipelines: All files can now be processed by a custom NextFlow pipeline, providing an easy-to-use API for any long-running downstream processing.
• Utilisation of AWS CDK for Infrastructure Deployment: Streamlining infrastructure setup and deployment using AWS Cloud Development Kit (CDK), ensuring rapid and reliable deployment of services. 
• Adoption of Cloud Security Best Practices: Implementing robust security measures following AWS security workshops, ensuring data integrity and compliance with regulatory standards. 
• Infrastructure Optimisation: Transitioning from AWS Elastic Beanstalk to AWS ECS (Elastic Container Service) to improve scalability and operational efficiency. Additionally, enhancing security measures with AWS Identity Centre for AWS access management. 

The most impactful outcome of the CDI Impact Accelerator was gaining deep insights into AWS SageMaker capabilities, pivotal for future developments and scalability plans. The support from AWS Solution Architects, particularly Emily Richardson, was instrumental in advancing genomics processing pipelines and enhancing overall platform capabilities for transformative healthcare solutions. 

Impact and Future Outlook

Phenopolis Ltd has made significant strides through their participation in the ¹û¶³Ó°Ôº CDI Impact Accelerator, notably achieving secure and scalable integration of AI-Assisted Annotation VisionSense for TrialSense using AWS. This milestone is pivotal in bridging the gap between Clinical Trial Management Systems and medical research, positioning Phenopolis to innovate within the clinical trial landscape.  

They have also enhanced their understanding and utilisation of AWS SageMaker's and AWS Batch’s capabilities, bolstered by support from AWS Solutions Architects, in developing Machine Learning and genomics processing pipelines. Looking forward, Phenopolis aims to advance doctor-in-the-loop AI annotations, enable real-time model training from annotations, and integrate genomics processing pipelines. 

With growing interest from hospitals, particularly in AI-assisted annotations, TrialSense is poised to democratize access to advanced clinical trial capabilities. This will empower smaller research organisations to compete effectively on a global clinical trial arena, positioning TrialSense to become the world’s leading Clinical Trial Management System.