University Hospital Zurich deploys AI tumor cell content quantification to improve NGS sample selection accuracy in routine molecular pathology

Manual tumor cell content (TCC) assessment prior to next-generation sequencing (NGS) is time-consuming, resource-intensive, and subject to limited pathologist availability, which can delay molecular testing initiation. Manual TCC assessment is also plagued by low accuracy and reproducibility — a referenced study found pathologists overestimated tumor cell percentage in 38% of cases, potentially causing false negative results. Inaccurate TCC quantification can lead to failed sequencing and destruction of tissue, a critical problem for patients with limited tissue available.

USZ deployed PathAI's AISight® Dx CE-IVD digital pathology platform along with the AIM-TumorCellularity (AIM-TC) algorithm to support routine TCC quantification prior to NGS workflows. The AI provides a stable, quantitative TCC estimate to increase consistency and confidence in sample quality selection for molecular diagnostics. The selection followed what is described as the most comprehensive analytical evaluation of an AI-based TCC algorithm to date, with preliminary results presented at three major pathology congresses.

AIM-TC demonstrated strong and reliable performance throughout USZ's rigorous evaluation, with accuracy maintained even in difficult real-world samples. The collaboration marks one of the first implementations of an AI-based tool in daily pathology operations in Switzerland. Preliminary results have been presented at the 2024 European Congress of Pathology, the 2025 European Congress of Digital Pathology, and the 2025 Swiss Society of Pathology Annual Meeting, with a detailed scientific publication forthcoming.

• AI-based TCC quantification addresses a well-documented reproducibility gap in manual pathologist assessment, directly reducing the risk of failed NGS runs and tissue waste.
• Rigorous multi-conference analytical validation before deployment builds the institutional confidence necessary to integrate AI into regulated routine diagnostic workflows.
• Deploying AI at the sample-selection step (pre-NGS) creates upstream quality control that protects downstream molecular testing accuracy and turnaround time.