Related Experiment Video
Updated: Nov 6, 2025
Author Spotlight: 3D Scanning and Augmented Reality for Enhanced Cancer Surgery Communication
Published on: December 15, 2023
Domain adaptation and self-supervised learning for surgical margin detection
Alice M L Santilli1, Amoon Jamzad2, Alireza Sedghi2
1School of Computing, Queen's University, Ontario, Canada. 14amls@queensu.ca.
Purpose:
One in five women who undergo breast conserving surgery will need a second revision surgery due to remaining tumor. The iKnife is a mass spectrometry modality that produces real-time margin information based on the metabolite signatures in surgical smoke. Using this modality and real-time tissue classification, surgeons could remove all cancerous tissue during the initial surgery, improving many facets of patient outcomes. An obstacle in developing a iKnife breast cancer recognition model is the destructive, time-consuming and sensitive nature of the data collection that limits the size of the datasets.
Methods:
We address these challenges by first, building a self-supervised learning model from limited, weakly labeled data. By doing so, the model can learn to contextualize the general features of iKnife data from a more accessible cancer type. Second, the trained model can then be applied to a cancer classification task on breast data. This domain adaptation allows for the transfer of learnt weights from models of one tissue type to another.
Results:
Our datasets contained 320 skin burns (129 tumor burns, 191 normal burns) from 51 patients and 144 breast tissue burns (41 tumor and 103 normal) from 11 patients. We investigate the effect of different hyper-parameters on the performance of the final classifier. The proposed two-step method performed statistically significantly better than a baseline model (p-value < 0.0001), by achieving an accuracy, sensitivity and specificity of 92%, 88% and 92%, respectively.
Conclusion:
This is the first application of domain transfer for iKnife REIMS data. We showed that having a limited number of breast data samples for training a classifier can be compensated by self-supervised learning and domain adaption on a set of unlabeled skin data. We plan to confirm this performance by collecting new breast samples and extending it to incorporate other cancer tissues.
Insights
This study introduces a novel method using self-supervised learning and domain adaptation for iKnife REIMS data to improve breast cancer detection. The approach enhances initial tumor removal accuracy, reducing the need for revision surgeries.
Area of Science:
- Surgical Oncology
- Analytical Chemistry
- Machine Learning
Background:
- Breast-conserving surgery often requires revision due to residual tumor, impacting patient outcomes.
- The iKnife (Intelligent Knife) uses mass spectrometry to provide real-time margin assessment from surgical smoke metabolites.
- Developing accurate iKnife models for breast cancer is challenging due to limited, sensitive data collection.
Purpose of the Study:
- To develop a robust iKnife breast cancer recognition model despite data limitations.
- To leverage self-supervised learning and domain adaptation for improved iKnife data analysis.
- To reduce the incidence of positive margins and subsequent revision surgeries in breast cancer treatment.
Main Methods:
- Utilized self-supervised learning on weakly labeled data to extract general iKnife metabolite features.
- Applied domain adaptation to transfer knowledge from a more accessible cancer type (skin) to breast cancer data.
- Investigated hyper-parameter effects on classifier performance using datasets of skin and breast tissue burns.
Main Results:
- Achieved statistically significant improvements over baseline models (p < 0.0001).
- The two-step method demonstrated high performance with 92% accuracy, 88% sensitivity, and 92% specificity.
- Successfully applied domain transfer for iKnife REIMS data, compensating for limited breast sample size.
Conclusions:
- This is the first demonstration of domain transfer for iKnife REIMS data in cancer classification.
- Self-supervised learning and domain adaptation effectively overcome data limitations in training iKnife breast cancer classifiers.
- Future work includes validating performance with more breast samples and expanding to other cancer types.
