Breast cancer outcome prediction with tumour tissue images and machine learning
Problem
Predicting breast cancer patient outcomes from tumour tissue images alone, rather than relying solely on human expert assessment.
Why it matters
Accurate prognosis guides treatment decisions for breast cancer patients and could reduce reliance on subjective expert review.
What the AI did
A machine learning method was trained on tissue microarray images from 868 patients to classify patients into low or high digital risk score groups, then tested on 431 patients against human expert classification.
Results
Hazard ratio 2.10 (95% CI 1.33-3.32, p=0.001) for breast cancer-specific survival; DRS remained an independent predictor in a nationwide cohort (N=1299).
Impact
Validated on a large nationwide cohort as an independent survival predictor, though clinical adoption is not stated.
Record
| Venue | Breast Cancer Research and Treatment |
|---|---|
| Published | 2019 |
| Work type | article |
| Authors | Riku Turkki, Dmitrii Bychkov, Mikael Lundin, Jorma J Isola, Stig Nordling, Panu E. Kovanen, Clare L. Verrill, Karl von Smitten et al. |
| Citations | 130 (OpenAlex, at import time) |
| References | 53 |
| Model or method | not stated in source |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1007/s10549-019-05281-1 |
| Open access | hybrid |
| Keywords | breast cancer, prognosis, machine learning, tissue microarray, survival prediction |
Importance rating
significant — thousands to millions affected
Breast cancer prognosis guides treatment decisions; ML from tissue images could reduce reliance on subjective expert review; validated on a nationwide cohort of 1299 patients.