Objective:

To compare pathologist review time and diagnostic performance with standard digital pathology versus AI-assisted digital pathology for breast biopsies, and to estimate potential effects on laboratory turnaround time.

Approach:

• Paired-reader study: Two experienced pathologists reviewed 194 breast biopsy cases using both workflows in a randomized order, with a 2-week washout period.
• Additional analysis: Researchers reviewed historical turnaround times for 3,298 cases to model the potential effect of AI-guided immunohistochemistry (IHC) preordering.

Key Findings:

• Mean review time fell from 71.12 seconds with standard digital pathology to 40.18 seconds with AI assistance, a reduction of about 44%.
• Cancer diagnostic agreement was 97% with standard review and 99% with AI assistance; sensitivity was 98% and 100%, respectively, and specificity was 95% and 98%.
• For DCIS, agreement was 93% without AI and 99% with AI, while sensitivity was 86% and 100%; the subgroup included only 21 DCIS-positive cases.
• The historical-data model estimated that AI-guided IHC preordering could reduce turnaround time by about 2.5 days for applicable true-positive cases; this was not tested prospectively.

Interpretation:

In this study, AI assistance was associated with shorter review times and high diagnostic agreement. The estimated turnaround-time reduction was modeled rather than demonstrated in clinical implementation.

Limitations:

• Single-center study involving only two pathologists.
• One representative slide per case was assessed, and cases with inadequate image quality were excluded.
• The study was not powered to establish differences in diagnostic accuracy; the small DCIS subgroup limits precision.
• The proposed turnaround-time benefit requires prospective evaluation.

Conclusion:

The authors stated that larger multicenter studies are warranted to confirm the findings.

Sources:

• Journal of Pathology Informatics (study source named in article context)