Monitoring Pathological Complete Response in Breast Cancer Patients Using Deep Learning
Author Information
Author(s): Maria Colomba Comes, Annarita Fanizzi, Samantha Bove, Luca Boldrini, Agnese Latorre, Deniz Can Guven, Serena Iacovelli, Tiziana Talienti, Alessandro Rizzo, Francesco Alfredo Zito, Raffaella Massafra
Primary Institution: I.R.C.C.S. Istituto Tumori “Giovanni Paolo II” Bari, Italy
Hypothesis
Can an ensemble deep learning model using Vision Transformers predict and monitor pathological complete response to neoadjuvant chemotherapy in breast cancer patients?
Conclusion
Changes in DCE-MRI at pre- and mid-treatment could affect the accuracy of predicting pathological complete response to neoadjuvant chemotherapy.
Supporting Evidence
- 37.2% of patients achieved pathological complete response.
- The model achieved an AUC of 91.4% on the internal test set.
- Statistically significant differences in AUC values were found when comparing the ViT model to competing models.
- The model was validated on an independent test set of 20 patients from the I-SPY2 trial.
Takeaway
Doctors can use special computer models to see if breast cancer treatment is working by looking at pictures of the tumors before and during treatment.
Methodology
The study used DCE-MRI scans from 86 breast cancer patients to develop a deep learning model that predicts pathological complete response based on imaging data.
Potential Biases
The study may be limited by the specific characteristics of the patient cohort and the imaging techniques used.
Limitations
The model's performance varied between the internal cohort and the independent test set, indicating potential context-dependency.
Participant Demographics
The cohort included 86 breast cancer patients, with 37.2% achieving pathological complete response.
Statistical Information
P-Value
0.0045
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website