Large-scale variability study: Robust AI tools for clinical IHC-quantification work outof- the-box

---

Anil Berger

Mindpeak GmbH, Germany

: Arch Clin Pathol

Abstract

---

Implementation of AI supported diagnostics is a key driver for establishing digital pathology in routine laboratories. Image analysis is particularly challenging for daily routine diagnostics as laboratories are using multiple types of pre-processing steps, IHC stainers and slide scanners. At Mindpeak, we developed a robust plug-and-play solution for immunohistochemical quantification in breast cancer samples for estrogen receptor (ER), progesterone receptor (PR) and Ki-67 as proliferation marker. Our solution provides reliable results across varying conditions. We show data from our recent large-scale study on IHC quantification in a real world setting, including 3 staining systems, 6 whole-slide scanners and cameras and 10 pathologists. The statistical analysis shows that our model is robust even under challenging conditions and ensures patient safety (non-inferiority margin of 0.05). We present results on how this model has been implemented and is now used daily in the routine workflows of several large routine laboratories.

Biography

---

Anil Berger is an experienced team leader and team player well at home in a cross-functional environment. Through effective P&Lmanagement I continuously grew my area's revenues by >10% p.a. With a structured way of working, attention to details and a strong affinity for figures and KPIs, I discover potential for optimization and translate them into continuous improvement processes for practice. Though motivation of my team and by living hands-on-mentality myself I successfully implement the company strategy so we can live up to our customers' expectations as well as company goals. My genuine customer orientation helped make Key Account Management successful and my market insight aided Business Development as well as the targeted broadening of our offer.