2020 Fiscal Year Research-status Report
Removing the Burden of Data Labeling: Automatic Surgical Video Understanding with Unsupervised Learning
| Project/Area Number |
20K23343
|
|---|
| Research Institution | Osaka University |
|---|
Principal Investigator |
李 良知 大阪大学, データビリティフロンティア機構, 特任研究員(常勤) (10875545)
|
|---|
| Project Period (FY) |
2020-09-11 – 2022-03-31
|
| Keywords | Few-shot Learning / Semantic Segmentation / Surgical Videos |
|---|
| Outline of Annual Research Achievements |
As most of the existing automatic surgical video analysis models require a large number of manually labeled data for training, this project aims to design a learning method to perform spatial and temporal segmentations with smaller requirements of humans’ input. In FY 2020, I mainly studied the following sub-topics towards the goal.
1.Surgical images/frames analysis using very few training samples. I developed an explainable few-shot learning method to give accurate recognition labels (as well as the explanations) to the input samples, which is very important for risk-sensitive areas like medicine. This work is presented at CVPRW 2021.
2.Surgical images/frames semantic segmentation in a weakly-supervised way. I developed a new training strategy for video semantic segmentation models to utilized unlabeled data to improve their segmentation performance. This work is published in IEEE Access.
3.Surgical videos temporal analysis using no labels. I developed a retrieval-based method to automatically predict surgical duration. This work is under submission.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
I planned to study the following sub-topic in FY 2020 and finished it as scheduled.
1.Unsupervised/weakly-supervised surgical semantic segmentation.
In addition, I also get some results from other two sub-topics.
1.Medical (responsible) few-shot learning.
2.Surgical temporal analysis.
|
| Strategy for Future Research Activity |
1.Unsupervised/weakly-supervised surgical video temporal segmentation. I plan to extend the duration prediction work to the temporal segmentation.
2.A prototype system to conduct automatic semantic and temporal segmentation.
|
| Causes of Carryover |
Reasons for Incurring Amount to be Used Next Fiscal Year: Many international conferences have been cancelled due to the pandemic. So the fee for travel and registration are not used.
Usage Plan: In the next year, the funding will be used for buying the workstation and other hardwares to build the prototype surgical analysis system, and the registration fee of the conferences that will be held online next year.
|
