A morphometric signature to identify ductal carcinoma in situ with a low risk of progression

Marcelo Sobral-Leite ¹, Simon P Castillo ², Shiva Vonk ¹

¹Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, The Netherlands.
²Division of Pathology and Laboratory Medicine, Department of Translational Molecular Pathology, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
³Division of Molecular Pathology, Oncode Institute, The Netherlands Cancer Institute, Amsterdam, Netherlands.
⁴Sarcoma Molecular Pathology Team, The Institute of Cancer Research, London, UK.
⁵Institute of Pathology and Medical Genetics, University Hospital Basel, Basel, Switzerland.
⁶Biostatistics Centre and Division of Psychosocial Research and Epidemiology, Netherlands Cancer Institute, Amsterdam, The Netherlands.
⁷Department of Surgery, Duke University Comprehensive Cancer Center, Durham, NC, USA.
⁸Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, The Netherlands. j.wesseling@nki.nl.
⁹Department of Pathology, Netherlands Cancer Institute - Antoni van Leeuwenhoek, Amsterdam, The Netherlands. j.wesseling@nki.nl.
¹⁰Department of Pathology, Leiden University Medical Center, Leiden, The Netherlands. j.wesseling@nki.nl.

⁰Division of Molecular Pathology, Netherlands Cancer Institute, Amsterdam, The Netherlands.

Npj Precision Oncology +

|

January 28, 2025

View abstract on PubMed

Summary

Artificial intelligence can analyze ductal carcinoma in situ (DCIS) tissue to predict progression to invasive breast cancer. This tool may help identify harmless DCIS, potentially reducing overtreatment for many women.

Area Of Science

Oncology
Pathology
Biomedical Engineering

Background

Ductal carcinoma in situ (DCIS) has variable progression rates to invasive breast cancer.
Current treatment paradigms for DCIS may lead to overtreatment for some patients with non-progressive disease.

Purpose Of The Study

To develop and validate an artificial intelligence (AI)-based morphometric analysis pipeline (AIDmap) for predicting DCIS progression.
To identify specific morphometric features associated with the risk of developing ipsilateral invasive breast cancer (iIBC).

Main Methods

Development of an AI pipeline (AIDmap) for analyzing morphometric features from digitized hematoxylin-eosin (H&E) stained tissue sections of pure primary DCIS.
Analysis of 689 DCIS H&E slides, with 226 cases progressing to iIBC and 463 not progressing.
Utilized ridge regression with cross-validation to predict 5-year iIBC-free survival.

Main Results

The AIDmap pipeline analyzed 55 morphometric variables derived from 15 duct morphological measurements.
The AI classifier achieved an area under the curve of 0.67 (95% CI 0.57-0.77) for predicting 5-year iIBC-free survival.
A combined clinical-morphometric signature (small ducts, low cell count, low DCIS/stroma ratio) was significantly associated with outcome (HR = 0.56).

Conclusions

The AIDmap system shows potential for identifying DCIS cases that are unlikely to progress to iIBC.
This AI-driven approach may help personalize treatment decisions and avoid overtreatment for women with non-progressive DCIS.