Posted in

AI Reclassifies Grade 2 Breast Tumors, Guiding Treatment Decisions

•Posted inRadiology

The Nottingham Histologic Grade (NHG) helps determine prognosis and treatment for breast cancer. However, NHG Grade 2 (NHG2) tumors are biologically heterogeneous, often leading to potential under- or over-treatment for patients in this intermediate-risk group. This significant challenge highlights the clinical need for improved NHG2 risk stratification. A new deep learning model, DeepRadGrade (DRG), offers a solution by using routine dynamic contrast-enhanced (DCE) MRI to stratify these intermediate-grade tumours effectively.

The Rationale for AI-Driven Grading

NHG is a vital prognostic factor. About 50% of breast cancer cases, however, fall into the ambiguous Grade 2 category. Conventional grading also suffers from inherent inter-observer variability. Furthermore, the intermediate risk in NHG2 tumours complicates treatment decisions, especially regarding neoadjuvant chemotherapy. Therefore, clinicians must identify which NHG2 tumours behave like Grade 1 (low-risk) and which are closer to Grade 3 (high-risk). Deep learning models successfully address this issue using digital whole slide histopathology images (WSIs). By contrast, the DRG model offers an alternative solution, leveraging standard DCE-MRI data instead.

Deep Learning for NHG2 Risk Stratification: DRG Findings

The DeepRadGrade (DRG) model was trained to distinguish between NHG1 and NHG3 tumours using DCE-MRI data from two large patient cohorts. After validation, the model applied its learned classification patterns to 456 NHG2 tumours. It segregated them into two new, distinct groups: DRG2- (NHG1-like, lower risk) and DRG2+ (NHG3-like, higher risk). The model demonstrated robust performance, achieving an AUC of 0.84 in the training phase. Importantly, the DRG classification showed clinical significance. Patients with DRG2+ tumours experienced significantly worse Recurrence-Free Survival (RFS) compared to those in the DRG2- group. The adjusted hazard ratio for recurrence was 2.39 (95% CI: 1.29-4.45), independent of standard prognostic factors like age and tumor stage. Moreover, incorporating the DRG classification substantially improved the overall prognostic accuracy of the Cox model.

Clinical Impact and Cost-Effectiveness

This new stratification method provides immediate clinical relevance. Consequently, doctors gain a reliable tool to individualise risk assessment for intermediate-grade breast cancer patients. Identifying the DRG2+ subgroup means clinicians can target these high-risk patients for intensive treatment, such as neoadjuvant chemotherapy. Conversely, the DRG2- group may be safely considered for de-escalation of therapy, minimising the side effects of unnecessary treatment. This deep learning approach uses routine DCE-MRI, making it a potentially cost-effective and highly accessible tool compared to expensive and time-consuming molecular or gene expression profiling assays. Therefore, this innovation promises to enhance therapeutic tailoring and reduce both over- and under-treatment in the clinic.

Frequently Asked Questions

Q1: What is the main challenge with Nottingham Histologic Grade 2 (NHG2) tumours?

NHG2 tumors represent an intermediate-risk category that is biologically heterogeneous, meaning some patients have a prognosis closer to low-risk (NHG1) and others closer to high-risk (NHG3). This heterogeneity makes treatment decisions, particularly regarding chemotherapy, difficult.

Q2: How does the DeepRadGrade (DRG) model reclassify NHG2 tumours?

The DRG model, a deep learning tool, analyses routine Dynamic Contrast-Enhanced (DCE) MRI scans. It reclassifies NHG2 tumors into two prognostic subgroups: DRG2- (NHG1-like, lower risk) and DRG2+ (NHG3-like, higher risk).

Q3: What are the clinical benefits of this advanced risk stratification?

The clinical benefits are individualised treatment and reduced over- or under-treatment. Patients with the high-risk DRG2+ subgroup may be identified for more aggressive therapy, while those in the lower-risk DRG2- group may be candidates for therapy de-escalation, minimising adverse side effects.

References

  1. Hadidchi R et al. A deep learning framework to stratify Nottingham histologic grade 2 breast tumours based on dynamic contrast-enhanced MRI. Eur Radiol. 2025 Dec 17. doi: 10.1007/s00330-025-12208-6. PMID: 41405689.
  2. Wang Y, Wählby C, Lundeberg J, Hartman J. Improved breast cancer histological grading using deep learning. Ann Oncol. 2021 Nov;32(11):1378-1386.
  3. Steiner DF, Silva K, Aldape K, et al. Deep learning models for histologic grading of breast cancer and association with disease prognosis. npj Breast Cancer. 2022 Oct 4;8(1):113.
  4. Wetstein SC, Mielke C, van der Laak J, et al. Validation of an AI-based solution for breast cancer risk stratification using routine digital histopathology images. NPJ Breast Cancer. 2024;10(1):1.