Back to all papers

Deep learning prediction of chemo-immunotherapy response using tumor perfusion ultrasound images.

June 12, 2026pubmed logopapers

Authors

Dimou K,Alexandrou F,Roussakis Y,Zamboglou C,Stylianopoulos T,Voutouri C

Affiliations (6)

  • Cancer Biophysics Laboratory, Department of Mechanical and Manufacturing Engineering, University of Cyprus, Nicosia, Cyprus.
  • Department of Research and Innovation, German Oncology Center, European University Cyprus, Limassol, Cyprus.
  • AnaBioSi-Data Ltd., Nicosia, Cyprus.
  • Department of Medical Physics, German Oncology Center, European University Cyprus, Limassol, Cyprus.
  • German Oncology Center, European University Cyprus, Limassol, Cyprus.
  • Department of Radiation Oncology, University of Freiburg - Medical Center, Freiburg, Germany.

Abstract

Tumor heterogeneity poses a significant challenge for predicting responses to cancer therapy, highlighting the need for the development of biomarkers to guide personalized treatment. Contrast-enhanced ultrasound (CEUS) imaging is an established method to assess tumor perfusion, which directly affects drug delivery and therapeutic efficacy, as poorly perfused tumors often limit the penetration of chemo- and immunotherapeutics. We developed a deep learning framework using CEUS imaging to predict the response of tumors to chemo-immunotherapy in murine models of breast cancer, fibrosarcoma, and melanoma. A convolutional neural network (CEUS-CNN) was trained on a dataset of 587 pre-treatment CEUS images to classify tumors as responsive, stable, or non-responsive based on RECIST version 1.1 (Response Evaluation Criteria in Solid Tumors) criteria (175 responsive cases, 136 stable, and 276 non-responsive). Additionally, synthetic data were created for the responsive and stable classes to address class disparity. Our framework attained an overall test accuracy of 0.877 (0.941 for responsive, 0.615 for stable, 0.963 for non-responsive) using only real data. The addition of synthetic data led to improved model performance, with a notable impact on the previously underperforming stable class. Our strategy enhanced the predictive capability of our model, raising the average test accuracy to 0.930 (1.000 for responsive, 0.769 for stable, 0.963 for non-responsive). These findings support CEUS imaging as a possible imaging biomarker of response to cancer therapy and further indicate that the incorporation of synthetic data can enhance model effectiveness, particularly for underrepresented classes. Together, they highlight the potential value of integrating AI with CEUS for personalized cancer treatment strategies.

Topics

Journal Article

Ready to Sharpen Your Edge?

Subscribe to join 11k+ peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.