Paul Mesange began his presentation by introducing himself as the head of the solids department at LabCorp and outlined the purpose of his talk: to explore the role of pixels in image analysis within clinical trials. Mesange described the lengthy and costly process of drug development, noting that clinical trials could span over a decade. He suggested that digital innovations, particularly in image analysis, had the potential to reduce both time and expense.
Mesange explained the traditional workflow for histology samples in clinical trials, which involves collecting samples from patients, shipping them to central labs, and having pathologists examine slides under microscopes. He highlighted that, since 2017, this process had shifted towards digital pathology, with slides now being scanned and transformed into images for analysis. This change enables remote collaboration and faster turnaround times, as images could be reviewed by pathologists worldwide, regardless of time zones.
He emphasised the growing complexity of clinical trials and the increasing demand for histology slide scanning, which had been rising annually. Mesange illustrated how digital image analysis, down to the pixel level, allowed for the identification of intricate patterns and supported a range of applications, from cancer detection to prognosis and treatment design. He noted that artificial intelligence (AI) models had demonstrated high diagnostic accuracy, often surpassing human pathologists, especially when used in conjunction with expert review.
Mesange discussed the advantages of image management systems, which not only facilitated collaboration and reporting but also integrated AI tools - either developed in-house or externally sourced. These systems improve patient care, reduce logistical burdens, and enable second opinions and consultations at any time.
He provided examples of AI-driven efficiencies, such as reducing annotation times for pathologists and supporting multiplex assays in spatial biology. Mesange also introduced the concept of virtual staining, where AI generated digital stains on slides, further streamlining workflows.
Regulatory considerations were addressed, with Mesange noting the importance of validation and oversight, ensuring that AI complemented rather than replaced pathologists. He concluded by reflecting on the future of digital pathology, highlighting the potential for multimodal AI and digital twins to advance personalised medicine, and emphasised the importance of using AI as a trusted partner in clinical trials to ultimately improve patient outcomes.
