Introduction: Immunohistochemical assessment of HER2, ER, PR, and Ki67 is essential in diagnosing invasive breast carcinomas and guides treatment, prognosis, and patient management. Here, we show how Indica Lab’s clinical software can be utilised to undertake research projects and train medical students at the University of Bern; focussing on a recent study comparing breast cancer IHC scoring by undergraduates to that of trained pathologists, when unassisted or assisted by HALO Breast IHC AI.

Fifty cases (250 slides) stained for ER, PR, HER2, and Ki67 were scored twice by three students and two certified pathologists. Initial scores were obtained using manual digital pathology methods (Visual Dx), followed by reassessment with AI assistance by HALO Breast IHC AI after a 4-week washout period (AI-Assisted Dx). Accuracy and agreement of the student scores were compared to those of the pathologists.

AI-Assisted Dx produced greater agreement at the clinical cutoff four all four biomarkers for both students (ER: 85% to 91%, PR: 19% to 73%, HER2: 47% to 82%, Ki67: 58% to 98% agreement) and pathologists (ER: 98% to 100%, PR: 44% to 90%, HER2: 71% to 100%, Ki67: 75% to 98% agreement). Similarly, Fleiss’ Kappa for all biomarkers increased for AI-Assisted Dx by an average of 0.42 for both students and pathologists. When both groups were combined, agreement at the clinical cutoff also showed improvement (ER: 85% to 91%, PR: 20% to 57%, HER2: 35% to 80%, Ki67: 51% to 98% agreement) indicating that student scores more closely aligned with pathologists when assisted by HALO Breast IHC AI.

HALO Breast IHC AI provides support for both students and certified pathologists, improving agreement and consensus scores across groups with different levels of experience. HALO Breast IHC AI can objectively standardise scores and be an effective training tool for students.