Weakly-Supervised Simultaneous Evidence Identification and Segmentation for Automated Glaucoma Diagnosis
Authors: Rongchang Zhao, Wangmin Liao, Beiji Zou, Zailiang Chen, Shuo Li809-816
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results show that our proposed WSMTL effectively and simultaneously delivers evidence identification, optic disc segmentation (89.6% TP Dice), and accurate glaucoma diagnosis (92.4% AUC). |
| Researcher Affiliation | Academia | Rongchang Zhao,1,2 Wangmin Liao,1,2 Beiji Zou,1,2 Zailiang Chen,1,2 Shuo Li3 1 School of Information Science and Engineering, Central South University, Changsha, China 2 Hunan Engineering Research Center of Machine Vision and Intelligent Medicine, Changsha, China 3 University of Western Ontario, London, ON, Canada |
| Pseudocode | Yes | Algorithm 1: Input: Evidence map EM Output: Optic disc mask STEP 1: Generate initial pixel patches 4: Generating the pixels patches X = x1, x2, ..., xn by applying k-means. 5: STEP 2: Similarity function learning 6: Learning the similarity function F(xp, xq) for all the patch pairs using the network proposed in (Zagoruyko and Komodakis 2015); 7: STEP 3: CCB optimization 8: Optimizing the CCB using the clustering loss L(xp, xq) as equation 8. 9: STEP 4: Reference of Optic disc mask 10: Input a test evidence map , forward propagate the data 11: through the CCB with trained weights, and get outputs for cluster assignment. Using ellipse fitting to obtain the optic disc segmentation mask. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code or links to a code repository. |
| Open Datasets | Yes | Our WSMTL is validated with the challenging dataset ORIGA650 (Cheng et al. 2017) with 168 glaucomatous and 482 normal eyes. |
| Dataset Splits | No | The paper states: 'The 650 images with manual labeled optic disc mask are randomly divided into 325 training images (Trainset, including 73 glaucoma cases) and 325 testing images (Testset, including 95 glaucoma).' It provides train and test splits but no explicit validation split from the dataset. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory used for running the experiments. |
| Software Dependencies | No | The paper mentions using DenseNet and pre-training on ImageNet, and that training uses stochastic gradient descent, but it does not specify version numbers for any programming languages, libraries, or frameworks (e.g., Python version, TensorFlow/PyTorch version, CUDA version). |
| Experiment Setup | No | The paper describes aspects of the model architecture (e.g., '4 dense blocks', 'number of output channels of the three scales to 6, 6, 12 and 24') and mentions 'stochastic gradient descent' as the optimizer. However, it does not provide specific hyperparameters critical for reproducibility, such as learning rate, batch size, or number of training epochs. |