Abstract
The Oncotype DX Recurrence Score (RS) guides adjuvant
chemotherapy decisions in ER+/HER2− early-stage breast cancer, but
genomic testing is not universally available. Ki67 reflects tumor proliferation
and is biologically associated with RS, yet routine assessment
is limited by interobserver variability. We investigated whether independently
trained and analytically validated AI-based Ki67 quantification
tools provide incremental information beyond routinely available clinicopathological
variables for estimating the genomic recurrence score.
We examined 167 patients who underwent clinical Oncotype DX testing.
Archival tumor tissue was assembled into tissue microarrays, stained for
Ki67, and digitized. Three independently developed AI models were analytically
validated against reference assessments before RS modeling.
Associations with continuous RS were evaluated using linear regression,
and logistic regression with cross-validation was used to identify patients
unlikely to have RS>25. AI-based Ki67 indices were associated with RS
and independently explained one-third of RS variance. When combined
with routine clinicopathological variables (age, histological grade, ER,
and PR), over half of RS variance was explained, and a 0.90 AUROC
was achieved for identifying RS>25. At a threshold of ≥95% sensitivity,
the combined variables ruled out 44% of patients, with an NPV of 0.97,
compared with 10% and 0.88, respectively, using routine clinicopathological
variables alone. Rather than replacing genomic testing, the combined
use of AI-based Ki67 quantification and routine clinicopathological variables
may serve as an accessible, scalable triage tool to identify patients
unlikely to have a high genomic recurrence score. Prospective external
validation in whole-slide or biopsy cohorts is required before clinical implementation.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/COMPAYL_075.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
BibTex
@InProceedings{DyAma_AIBased_MICCAISAT2026,
author = { Dy, Amanda AND Shafique, Abubakr AND Qin, Xiaoli AND Androutsos, Dimitrios AND Done, Susan J. AND Khademi, April},
title = { { AI-Based Ki67 Quantification for Triage of Genomic Recurrence Score in Breast Cancer } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17251},
month = {pending},
page = {pending}
}