Context Matters: Graph-based Self-supervised Representation Learning for Medical Images
Authors: Li Sun, Ke Yu, Kayhan Batmanghelich4874-4882
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on large-scale Computer Tomography (CT) datasets of lung images show that our approach compares favorably to baseline methods that do not account for the context. We evaluate the performance of the proposed model on two large-scale datasets of 3D medical images. We compare our model with various baseline methods, including both supervised approaches and unsupervised approaches. |
| Researcher Affiliation | Academia | Li Sun , Ke Yu , and Kayhan Batmanghelich University of Pittsburgh, USA {lis118, key44, kayhan}@pitt.edu |
| Pseudocode | Yes | Algorithm 1 Interleaving update algorithm Require: Conditional encoder E( , ), GCN G( , ). Input: Image patch xj i, anatomical landmark pj, adjacency matrix Ai. for step t = 1, Tmax do for step tl = 1, Tl do Randomly sample a batch of Bl subjects for j = 1, N do hj i E(aug(xj i), pj) Update E by backpropagating Ll end for end for for step tg = 1, Tg do Randomly sample a batch of Bg subjects Si G(concat({E(aug(xn i ), pn)}N n=1), Ai) Update G by backpropagating Lg end for end for |
| Open Source Code | Yes | Our code and supplementary material are available at https://github.com/batmanlab/Context Aware SSL |
| Open Datasets | Yes | COPDGene Dataset COPD is a lung disease that makes it difficult to breathe. The COPDGene Study (Regan et al. 2011) is a multi-center observational study designed to identify the underlying genetic factors of COPD. We use a large set of 3D thorax computerized tomography (CT) images of 9,180 subjects from the COPDGene dataset in our study. Mos Med Dataset We use 3D CT scans of 1,110 subjects from the Mos Med dataset (Morozov et al. 2020) provided by municipal hospitals in Moscow, Russia. COVID-19 CT Dataset The combined dataset has 80 subjects, in which 35 positive subjects are from multiple publicly available COVID-19 datasets (Jun et al. 2020; Bell 2020; Zhou et al. 2020), and 45 healthy subjects randomly sampled from the LIDC-IDRI dataset (Armato III et al. 2011) as negative samples. |
| Dataset Splits | Yes | We report average R2 scores with standard deviations in five-fold cross-validation. We use the average test accuracy in five-fold cross-validation as the metric for quantifying prediction performance. |
| Hardware Specification | No | The paper states 'The experiments are performed on 2 GPUs, each with 16GB memory.' This does not specify the exact GPU model or CPU details. |
| Software Dependencies | No | The paper mentions software like ANTs and lungmask but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | We train the proposed model for 30 epochs. We set the learning rate to be 3 10 2. We also employed momentum = 0.9 and weight decay = 1 10 4 in the Adam optimizer. The patch size is set as 32 32 32. The batch size at patch level and subject level is set as 128 and 16, respectively. We let the representation dimension F be 128. The number of negative samples during training is set as 4096. The data augmentation includes random elastic transform, adding random Gaussian noise, and random contrast adjustment. The temperature τ is chosen to be 0.2. |