Author: Xuehai He; Xingyi Yang; Shanghang Zhang; Jinyu Zhao; Yichen Zhang; Eric Xing; Pengtao Xie
Title: Sample-Efficient Deep Learning for COVID-19 Diagnosis Based on CT Scans Document date: 2020_4_17
ID: l3f469ht_74
Snippet: In this paper, we study how to develop sample-efficient deep learning methods to accurately diagnose COVID-19 from CT scans. To facilitate the open research in this area, we build COVID19-CT, a dataset containing 349 CT scans positive for COVID-19. To our best knowledge, it is the largest COVID19-CT dataset that is publicly available to date. Though the largest, it still incurs a high risk of overfitting for data-hungry deep learning models. To r.....
Document: In this paper, we study how to develop sample-efficient deep learning methods to accurately diagnose COVID-19 from CT scans. To facilitate the open research in this area, we build COVID19-CT, a dataset containing 349 CT scans positive for COVID-19. To our best knowledge, it is the largest COVID19-CT dataset that is publicly available to date. Though the largest, it still incurs a high risk of overfitting for data-hungry deep learning models. To reduce this risk, we develop data efficient methods that are able to mitigate data deficiency. We propose Self-Trans, a self-supervised transfer learning approach that learns expressive and unbiased visual feature representations that are robust to overfitting. Through extensive experiments, we demonstrate the effectiveness of our methods. is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.13.20063941 doi: medRxiv preprint
Search related documents:
Co phrase search for related documents- deep learning model and high risk: 1, 2, 3, 4, 5, 6
- deep learning model and method effectiveness: 1, 2
- feature representation and method effectiveness: 1, 2
- feature representation and transfer learning: 1, 2, 3, 4
- good knowledge and high risk: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36
- good knowledge and risk reduce: 1, 2, 3, 4, 5, 6, 7, 8, 9
- good knowledge and transfer learning: 1
- high risk and method effectiveness: 1, 2, 3, 4, 5, 6
- high risk and open research: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
- high risk and risk reduce: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75
- high risk and transfer learning: 1, 2, 3, 4, 5, 6, 7, 8
- method effectiveness and open research: 1, 2, 3
- method effectiveness and risk reduce: 1
- method effectiveness and transfer learning: 1, 2, 3
- method effectiveness demonstrate and open research: 1, 2
- method effectiveness demonstrate and transfer learning: 1
- open research and risk reduce: 1, 2, 3
- open research and transfer learning: 1, 2, 3, 4, 5
- risk reduce and transfer learning: 1, 2, 3, 4
Co phrase search for related documents, hyperlinks ordered by date