On April 8 and 9, a significant workshop on deep learning and big data will be held at the Downtown Library Complex, Room 104, organized by West Virginia University (WVU). This event is poised to attract attention as it seeks to address critical issues surrounding the deployment of artificial intelligence (AI) in clinical settings.
The workshop comes at a time when the pharmaceutical industry faces escalating costs, with the global average for Phase 3 development programs now exceeding $1.2 billion. This financial burden underscores the urgency for organizations to optimize their AI strategies and operational frameworks.
Recent surveys reveal that fewer than 12% of pharmaceutical organizations have implemented formal drift detection mechanisms for their production clinical AI models. This lack of monitoring is concerning, as it can lead to a significant gap between the potential value of clinical AI and its actual operational contributions. As one expert noted, “The result is a widening gap between the potential value of clinical AI and its realized operational contribution.”
Moreover, the average period between deployment and database lock for these Phase 3 programs is approximately 28 months, highlighting the need for more efficient processes. The FDA’s proposed Predetermined Change Control Plan framework aims to address these challenges by envisioning pre-approved protocols for how AI models can be updated in production.
Organizations that have begun deploying feature stores report a median 43% reduction in duplicated feature engineering efforts across model teams, indicating a path forward for improving efficiency. However, the productive deployment of AI in clinical data operations is contingent on the maturation of MLOps infrastructure, which applies the principles of DevOps to AI.
As the workshop approaches, participants are encouraged to engage with the pressing issues of continuous monitoring and drift detection. One expert cautioned, “Without continuous monitoring and drift detection, models degrade invisibly.” This highlights the critical need for organizations to ensure the accuracy, reliability, and defensibility of their AI models long after deployment.
As the deep learning workshop unfolds, it is expected to generate valuable insights and foster discussions on how to bridge the gap between AI’s potential and its practical applications in the pharmaceutical industry. The question remains: “Will the AI your organization deploys still be working accurately, reliably, and defensibly two years after deployment?”
