Disentangling Disease-related Representation from Obscure for Disease Prediction
Authors: Chu-Ran Wang, Fei Gao, Fandong Zhang, Fangwei Zhong, Yizhou Yu, Yizhou Wang
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results on one public dataset DDSM and three in-house datasets demonstrate that the proposed strategy can achieve state-of-the-art performance. |
| Researcher Affiliation | Collaboration | 1Center for Data Science, Peking University, Beijing, China 2Center on Frontiers of Computing Studies, School of Computer Science, Peking University, Beijing,China 3School of Computer Science, Peking University, Beijing, China 4School of Artificial Intelligence, Peking University, Beijing, China 5AI lab, Deepwise Healthcare, Beijing, China 6Department of Computer Science, The University of Hong Kong, Hong Kong 7Inst. for Artificial Intelligence, Peking University, Beijing, China. |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement or link for open-source code availability. |
| Open Datasets | Yes | To evaluate the effectiveness of our model, take mammogram mass benign/malignant classification as an example, we consider both the public dataset DDSM (Bowyer et al., 1996) and three in-house datasets (Inh1, Inh2, Inh3). |
| Dataset Splits | Yes | The details of training/validating/testing data we use for experiments are shown in Fig. 7. And the details of the number of the obscured/non-obscured masses in each dataset we use are shown in Fig. 6. All settings are the same as (Wang et al., 2021a) for fair comparison. [...] The number of ROIs in each dataset. Each histogram denotes a each dataset and each dataset is divided into training, validating and testing by 8:1:1. |
| Hardware Specification | No | The paper does not explicitly describe the specific hardware (e.g., GPU/CPU models, memory) used for running its experiments. |
| Software Dependencies | No | We implement all models with Py Torch. For a fair comparison, all methods are conducted under the same setting and share the same encoder backbone, i.e., Res Net34 (He et al., 2016). |
| Experiment Setup | Yes | We implement Adam to train our model. [...] share the same encoder backbone, i.e., Res Net34 (He et al., 2016). [...] For disentangle learning, we first use only composite obscured data to pretrain for 500 epochs while the number of samples is the same as the real dataset. Then we add composite obscured data to original real data for classification training, and the number of added composite data is 30% of the original datasets. [...] The inputs are resized into 224 224 with random horizontal flips and fed into networks. |