FAST: A Dual-tier Few-Shot Learning Paradigm for Whole Slide Image Classification
Authors: Kexue Fu, xiaoyuan luo, Linhao Qu, Shuo Wang, Ying Xiong, Ilias Maglogiannis, Longxiang Gao, Manning Wang
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on binary and multi-class datasets demonstrate that our proposed method significantly surpasses existing few-shot classification methods and approaches the accuracy of fully supervised methods with only 0.22% annotation costs. |
| Researcher Affiliation | Academia | Kexue Fu16 , Xiaoyuan Luo23 , Linhao Qu23 , Shuo Wang23, Ying Xiong4, Ilias Maglogiannis5, Longxiang Gao16 , Manning Wang23 1Key Laboratory of Computing Power Network and Information Security, Ministry of Education, Shandong Computer Science Center (National Supercomputer Center in Jinan), Qilu University of Technology (Shandong Academy of Sciences), Jinan, China 2Digital Medical Research Center, School of Basic Medical Sciences, Fudan University 3Shanghai Key Lab of Medical Image Computing and Computer Assisted Intervention 4 Fudan University 5 University of Piraeus 6 Shandong Provincial Key Laboratory of Computing Power Internet and Service Computing, Shandong Fundamental Research Center for Computer Science, Jinan, China |
| Pseudocode | No | The paper describes the method using textual descriptions and flowcharts (Figure 2) but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | All codes and models will be publicly available on https://github.com/fukexue/FAST. |
| Open Datasets | Yes | We evaluated our method on the public WSI datasets CAMELYON16 [4] and TCGA-RENAL [17]. |
| Dataset Splits | No | Finally, all WSIs were divided into training and testing sets in a ratio of 70% and 30%, respectively. |
| Hardware Specification | Yes | All models are trained and tested on an RTX 3090 GPU with 24GB memory. |
| Software Dependencies | No | FAST employs the image encoder and text encoder from the pre-trained CLIP-RN50 [41]. During training, we utilized the Adam optimizer with learning rates set as follows: 0.001 for the feature cache, 0.01 for the label cache, and 0.001 for the tokens in the prior branch. |
| Experiment Setup | Yes | We set the number of learnable tokens to 10. During training, we utilized the Adam optimizer with learning rates set as follows: 0.001 for the feature cache, 0.01 for the label cache, and 0.001 for the tokens in the prior branch. We fine-tune our model with batch size of 4096 for 20,000 epochs. |