81 Assessing Recurrence Likelihood in Hormone Receptor–Positive/HER2-Negative Breast Cancer Patients Directly From MRI Using Imaging AI

Publication
Article
Miami Breast Cancer Conference® Abstracts Supplement42nd Annual Miami Breast Cancer Conference® - Abstracts
Volume 39
Issue 4
Pages: 67

81 Assessing Recurrence Likelihood in Hormone Receptor–Positive/HER2-Negative Breast Cancer Patients Directly From MRI Using Imaging AI

81 Assessing Recurrence Likelihood in Hormone Receptor–Positive/HER2-Negative Breast Cancer Patients Directly From MRI Using Imaging AI

Background/Significance

Early-stage hormone receptor (HR)–positive, HER2-negative breast cancer is treated with endocrine therapy with or without chemotherapy. Commercially available genomic assays may not fully capture disease heterogeneity when performed on a single tissue block. Radiographic imaging with multi-modal AI allows for a global assessment of tumor heterogeneity. We developed and tested an AI-based model incorporating pretreatment dynamic contrast-enhanced (DCE)-MRI with clinicopathologic factors to predict recurrence risk in HR–positive/HER2-negative breast cancer.

Materials and Methods

A multi-institutional cohort of 1072 women diagnosed with invasive HR–positive/HER2-negative breast cancer who had follow-up and pretreatment DCE-MRI was identified. Institutions were assigned to either algorithm training or testing, resulting in datasets with 522 and 550 women, respectively. An AI risk model was developed that utilized clinicopathological data (age, race/ethnicity, T stage, N stage, grade) and image-based spatial features of the tumor and surrounding tissues derived from the DCE-MRI (Figure 1A, B). The risk model was calibrated on the training dataset to produce a score correlating with probability of loco-regional recurrence or distant recurrence (hereafter RP). High- and low-risk of recurrence categories were defined by optimizing a threshold in the training dataset. Finally, the risk model was assessed in the testing dataset using this threshold.

Results

In the test cohort, patients had a median follow-up time of 4.5 years, median age of 54 ± 11 years, and 15.6% were African American. The majority of women had T1-T2 disease (89.6%), and 34.4% had 1 to 3 involved lymph nodes. The 5-year RP for women predicted to be low-risk was 6.2% (95% CI, 3.8%-10.0%) compared with 17.4% (95% CI, 11.9%-25.2%) for women predicted to be high-risk (Figure 1C). The adjusted hazard ratio including clinicopathologic factors between low- and high-risk groups was 3.1 (95% CI, 1.7-5.8; P <.005). Similar hazard ratios between low- and high-risk were observed in important subgroups: node negative (3.6; 95% CI, 1.6-8.1), node positive (2.2; 95% CI, 0.8-5.9), age < 50 years (2.5; 95% CI, 1.0-6.2), and age > 50 years (2.9; 95% CI, 1.2-6.8).

Figure 1. A) AI model development includes whole-tumor imaging and clinicopathologic variables. B) Spatial characteristics correlate with recurrence risk. C) Results in the independent testing dataset.

Figure 1. A) AI model development includes whole-tumor imaging and clinicopathologic variables. B) Spatial characteristics correlate with recurrence risk. C) Results in the independent testing dataset.

Conclusion

Our AI-based prognostic tool incorporating imaging allows personalized treatment planning in real time in women with early-stage HR–positive/HER2-negative breast cancer. The prognostic benefit exceeded that of clinical features alone and was observed regardless of age and lymph node involvement.

Articles in this issue

2 Preventive Care and Screening Adherence Among Women Surviving Breast Cancer
2 Preventive Care and Screening Adherence Among Women Surviving Breast Cancer
3 Intraoperative Radiotherapy: Alive and Well in the Bronx
3 Intraoperative Radiotherapy: Alive and Well in the Bronx
4 Evaluating AI-Driven Responses in Breast Reconstruction: A Comparative Study of Response Formats
4 Evaluating AI-Driven Responses in Breast Reconstruction: A Comparative Study of Response Formats
5 Correlation Between Visual Impairment and Breast Cancer: A Cross-Sectional Study Based on the National Health Interview Surveys
5 Correlation Between Visual Impairment and Breast Cancer: A Cross-Sectional Study Based on the National Health Interview Surveys
6 Peer Support Programming Among Women At-Risk for Surviving Breast Cancer: Facilitators and Barriers to Community-Based Patient Navigation and the Role of Quality of Life
6 Peer Support Programming Among Women At-Risk for Surviving Breast Cancer: Facilitators and Barriers to Community-Based Patient Navigation and the Role of Quality of Life
7 Metaplastic Breast Cancer: A Retrospective Chart Review of Clinical Features
7 Metaplastic Breast Cancer: A Retrospective Chart Review of Clinical Features
8 A Case Series Exploring Characteristics and Outcomes of Metachronous Primary Breast and Lung Cancer in a Diverse Cohort
8 A Case Series Exploring Characteristics and Outcomes of Metachronous Primary Breast and Lung Cancer in a Diverse Cohort
9 Body Mass Index, Cancer Risk Behaviors, and Readiness for Dietary Change Among Women Surviving With Breast Cancer
9 Body Mass Index, Cancer Risk Behaviors, and Readiness for Dietary Change Among Women Surviving With Breast Cancer
10 AI as a Bridge: Can ChatGPT Help Patients Understand Their Breast Radiology Reports?
10 AI as a Bridge: Can ChatGPT Help Patients Understand Their Breast Radiology Reports?
12 Gut Microbiome Composition and Pathological Complete Response After Chemotherapy in Breast Cancer: Insights From a Pilot Study
12 Gut Microbiome Composition and Pathological Complete Response After Chemotherapy in Breast Cancer: Insights From a Pilot Study
13 Preliminary Analysis of Change During Treatment of Financial Toxicity and Quality of Life in Breast Cancer Patients
13 Preliminary Analysis of Change During Treatment of Financial Toxicity and Quality of Life in Breast Cancer Patients
15 Utilizing Circulating Tumor Cells to Guide HER2-Directed Therapy in IHC/FISH-Negative HER2+ Metastatic Breast Cancer
15 Utilizing Circulating Tumor Cells to Guide HER2-Directed Therapy in IHC/FISH-Negative HER2+ Metastatic Breast Cancer
16 A Miami Hospital’s Infrastructure to Help Decrease Late-Stage Breast Cancer Diagnosis and Improve Health Equity
16 A Miami Hospital’s Infrastructure to Help Decrease Late-Stage Breast Cancer Diagnosis and Improve Health Equity
17 Salmonella and the Breast: A Literature Review of Salmonella-Induced Breast Abscesses
17 Salmonella and the Breast: A Literature Review of Salmonella-Induced Breast Abscesses
18 Tolerability of First-Line Treatment With Ribociclib for Metastatic Breast Cancer Using 2 Large US Data Sources
18 Tolerability of First-Line Treatment With Ribociclib for Metastatic Breast Cancer Using 2 Large US Data Sources
Recent Videos
Patients with node-negative disease who are older and have comorbidities may not be suitable to receive CDK4/6 inhibitors.
An observed carryover effect with CDK4/6 inhibitors may reduce the risk of recurrence years after a patient stops treatment.
“If you have a [patient in the] fourth or fifth line, [JNJ-5322] could be a valid drug of choice,” said Rakesh Popat, BSc, MBBS, MRCP, FRCPath, PhD.
Earlier treatment with daratumumab may be better tolerated for patients with pretreated MRD-negative multiple myeloma.
The trispecific antibody JNJ-5322 demonstrated superior efficacy vs approved agents in multiple myeloma in results shared at the 2025 EHA Congress.
Despite CD19 CAR T-cell therapy exhibiting efficacy in patients with relapsed/refractory large B-cell lymphoma, less than half achieve long-term remission.