Non-Invasive AI Tool Predicts Breast Cancer Aggressiveness

Scientists have introduced a significant breakthrough in non-invasive Ki-67 prediction for breast cancer patients. Currently, evaluating Ki-67 expression requires invasive biopsies and specialized pathological analysis. Consequently, these procedures can be painful and may delay important treatment decisions. However, this new deep learning model uses routine ultrasound images to assess tumor markers accurately. By integrating clinical data with imaging features, the tool offers a safer and faster alternative for monitoring cancer progression.

The Impact of Non-invasive Ki-67 Prediction

Researchers developed this advanced model through a multicenter study involving four medical centers. They extracted detailed ultrasound features using the ResNet-50 architecture. Furthermore, the team combined these features with specific clinical information through logistic regression. The results showed that the model achieved an impressive area under the curve (AUC) of 0.828 in external tests. Therefore, this technology successfully identifies high-risk individuals without needing extra surgical interventions. This approach significantly enhances the diagnostic workflow for busy oncology clinics.

Better Risk Stratification for Indian Clinics

The study also examined how well the model predicts lymph node metastasis and overall survival rates. Interestingly, the high Ki-67 group identified by the AI displayed much higher rates of metastasis. Additionally, these patients faced significantly shorter recurrence-free survival times compared to the low-risk group. This means doctors can now use standard imaging to tailor treatment plans more effectively for their patients. Moreover, the visualization through nomograms helps clinicians understand the logic behind the AI\’s predictions. As a result, the model builds clinical trust through transparency and reliable performance.

Frequently Asked Questions

Q1: Why is Ki-67 prediction important for breast cancer patients?

Ki-67 is a protein that serves as a proliferation marker, indicating how quickly cancer cells are dividing. High levels often correlate with aggressive tumors and a higher risk of recurrence.

Q2: Can ultrasound replace a biopsy for measuring Ki-67?

While a biopsy is the traditional gold standard, this deep learning model provides a non-invasive way to predict Ki-67 levels with high accuracy, assisting in risk stratification when a biopsy is not immediately feasible.

References

Yu W et al. Deep learning model for noninvasive prediction of Ki-67 expression and prognostic stratification in breast cancer: a multicenter retrospective study. Eur Radiol. 2026 May 07. doi: 10.1007/s00330-025-12203-x. PMID: 42095876.
Chakraborty C et al. Indian scientists tap AI to identify aggressive breast cancer. The Economic Times. 2017 Jul 23.
Patil et al. Ki-67 biomarker in breast cancer of Indian women. PMC. 2011 Mar 15.