Datasets:

mfromm
/

AMSR

{"forum": "0W3J_Y-zlS", "submission_url": "https://openreview.net/forum?id=O6rt5JpTUl", "submission_content": {"title": "Calcium Score prediction and CAC localization using anisotropic convolutional networks.", "authors": ["Mario Viti", "Hugues Talbot", "Nicolas Gogin"], "authorids": ["mario.viti1@ge.com", "hugues.talbot@centralesupelec.fr", "nicoals.gogin@ge.com"], "keywords": ["CAC", "Calcium Score", "Cardiac CT", "Deep Learning", "Semantic Segmentation"], "TL;DR": "Calcium Score prediction and CAC localization using anisotropic convolutional networks.", "abstract": "Abstract\u2014In this work we propose to apply deep learning\nsemantic segmentation techniques to calcium quantification and\nlocalization. 3D CT chest imaging is an essential support to\ndiagnosis of cardiovascular disease and coronaires calcification\nburden is one of its strongest indicators. CAC is quantified using a\nper coronary branch Agatston score. In this clinical context using\ndeep learning techniques for multi class segmentation we could\ndesign an algorithm that automatically localizes and quantifies\ncalcifications volumetry and Agatston score. The architecture\nused is inspired by Vnet [7], a popular model adapted to this\nparticular CT exam modality, the key contribution is the use\nof anisotropic pooling and unpooling layers. 124 patients were\nprovided by ***** and manually annotated by experts\nwith clinical feedback. As a result we could achieve 0.9 average\nR2 = 1 - rMSE (relative mean square error) on multiple branches\non a test set of 14 patient left out from the whole dataset.\nIndex Terms\u2014CAC, Calcium Score, Cardiac CT, Deep Learning,\nSemantic Segmentation.", "pdf": "/pdf/77cc1e7e3f3773ff006f04da046ffd640d182dbe.pdf", "track": "short paper", "paper_type": "well-validated application", "paperhash": "viti|calcium_score_prediction_and_cac_localization_using_anisotropic_convolutional_networks", "_bibtex": "@misc{\nviti2020calcium,\ntitle={Calcium Score prediction and {\\{}CAC{\\}} localization using anisotropic convolutional networks.},\nauthor={Mario Viti and Hugues Talbot and Nicolas Gogin},\nyear={2020},\nurl={https://openreview.net/forum?id=O6rt5JpTUl}\n}"}, "submission_cdate": 1579955736154, "submission_tcdate": 1579955736154, "submission_tmdate": 1587172133291, "submission_ddate": null, "review_id": ["PTsJlx4p7", "6W456qUxK", "m4eWI-R5Li", "hve_qM2Bmp"], "review_url": ["https://openreview.net/forum?id=O6rt5JpTUl&noteId=PTsJlx4p7", "https://openreview.net/forum?id=O6rt5JpTUl&noteId=6W456qUxK", "https://openreview.net/forum?id=O6rt5JpTUl&noteId=m4eWI-R5Li", "https://openreview.net/forum?id=O6rt5JpTUl&noteId=hve_qM2Bmp"], "review_cdate": [1584143990484, 1584034473363, 1583325786828, 1582101766354], "review_tcdate": [1584143990484, 1584034473363, 1583325786828, 1582101766354], "review_tmdate": [1585229940558, 1585229940055, 1585229939549, 1585229939042], "review_readers": [["everyone"], ["everyone"], ["everyone"], ["everyone"]], "review_writers": [["MIDL.io/2020/Conference/Paper219/AnonReviewer4"], ["MIDL.io/2020/Conference/Paper219/AnonReviewer1"], ["MIDL.io/2020/Conference/Paper219/AnonReviewer3"], ["MIDL.io/2020/Conference/Paper219/AnonReviewer2"]], "review_reply_count": [{"replyCount": 0}, {"replyCount": 0}, {"replyCount": 0}, {"replyCount": 0}], "review_replyto": ["0W3J_Y-zlS", "0W3J_Y-zlS", "0W3J_Y-zlS", "0W3J_Y-zlS"], "review_content": [{"title": "Incomplete paper", "review": "There are a lot of grammatically wrong sentences and typos, so it was hard to follow. There\u2019s no conclusion or discussion. The readers may not get the point of this paper without conclusion. Overall, the paper is poorly organized. Figure 1 and 3 were not referenced in the manuscript. Please use distinct colors for different anatomy.", "rating": "1: Strong reject", "confidence": "4: The reviewer is confident but not absolutely certain that the evaluation is correct"}, {"title": "No clear contribution", "review": "Summary\nThe authors propose to use a V-Net for segmentation of coronary artery calcification (CAC) voxels in 3D chest CT images.\n\nStrengths\n-\tCalcium scoring is a clinically relevant task.\n\nWeaknesses\n-\tThere have been many calcium scoring papers using deep learning, it\u2019s unclear what the proposed method adds to those. The quantitative evaluation is very different from common evaluation approaches in this field and it is thus difficult to compare the obtained results to other methods. This could be addressed by evaluating in a public benchmark such as the orCaScore challenge (https://aapm.onlinelibrary.wiley.com/doi/abs/10.1118/1.4945696, https://orcascore.grand-challenge.org/).\n-\tAuthors evaluated their method on a test set of 14 patients, this is very small compared to other papers that have test sets containing hundreds of images (https://ieeexplore.ieee.org/document/8094970, https://ieeexplore.ieee.org/document/7061518, see https://www.frontiersin.org/articles/10.3389/fcvm.2019.00172/full for an overview). \n-\tThe authors should carefully revise the related works section, many statements about related papers are incorrect. E.g., Lessmann et al. did not combine contrast and non-contrast scans. Santini et al. and Huo et al. did not estimate calcium scores directly but performed segmentation. On the other hand, De Vos et al. (https://ieeexplore.ieee.org/abstract/document/8643342, not cited) did. The authors write that no previous methods have localized deposits within branches, but actually this is quite common. See e.g. the participating methods in https://aapm.onlinelibrary.wiley.com/doi/abs/10.1118/1.4945696. \n-\tTo address the problem of class imbalance between calcified voxels and background voxels, the authors dilate all lesions in the reference standard. However, this is likely to lead to oversegmentation. \n-\tThe paper is not well-prepared, all image captions seem to be the same. \n\nDetailed comments\n-\tQuite some typos and grammatical errors, please check carefully. E.g. \u2018coronaires\u2019, \u2018it is therefore become\u2019, \u2018Previous the deep learning era\u2019, \u2018simmetry\u2019, etc.\n-\tThe statement \u2018For such modality the most likely intensity of calcium is 130 HU on the Hounsfield scale\u2019 is incorrect, this is only a threshold. Density values can be much higher.\n-\tFig. 1 is not particularly useful for this paper as it does not show any coronary calcifications. \n-\tWhat do the authors mean with \u2018voxelometry\u2019?\n", "rating": "1: Strong reject", "confidence": "5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature"}, {"title": "Not a well-validated application of deep learning", "review": "This paper describes a method for automatic detection and labeling of coronary calcification in CT. The authors use a VNet with anisotropic down- and up-sampling to account for the lower resolution along the z axis that is typical for calcium scoring CT scans. This short paper claims to present a well-validated application of deep learning in medical imaging, but does unfortunately not live up to this claim. The method description is clear, but the data and annotation protocol are unclear, the test set is small (only 14 subjects, even though the amount of coronary calcification per patient is often rather small) and the paper contains many mistakes (the related work section confuses references, e.g., Yang et al. used non-contrast and contrast-enhanced CT scans, not Lessmann et al., who on the other hand predicted the location of the calcification, which the authors claim has not been done before; the captions of Figures 2-4 do not describe the figures, Figures 3 and 4 even have the exact same caption; the results section suddenly mentions numbers for \"aorta\", which has not been mentioned before; etc).", "rating": "1: Strong reject", "confidence": "5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature"}, {"title": "The paper exceeds the page limit and thereby violates the submission rules.", "review": "The MIDL 2020 author instructions (https://2020.midl.io/author-instructions.html) clearly state that \"Short papers are up to 3 pages (excluding references and acknowledgements)\". This requirement is not met by the submission.\n\nThe paper was submitted as \"well-validated application\", which is questionnable given the empirical validation. The Appendix looks unmotivated and unrelated to the text. The plots in Figure 2 contain JPEG artifacts and fail to communicate how well the prediction performs. Here, a Bland-Altman or similar plot would be better suited.", "rating": "1: Strong reject", "confidence": "5: The reviewer is absolutely certain that the evaluation is correct and very familiar with the relevant literature"}], "comment_id": [], "comment_cdate": [], "comment_tcdate": [], "comment_tmdate": [], "comment_readers": [], "comment_writers": [], "comment_reply_content": [], "comment_content": [], "comment_replyto": [], "comment_url": [], "meta_review_cdate": 1586217520857, "meta_review_tcdate": 1586217520857, "meta_review_tmdate": 1586217520857, "meta_review_ddate ": null, "meta_review_title": "MetaReview of Paper219 by AreaChair1", "meta_review_metareview": "The paper received clear feedback that it was unfinished and did not meet the criteria for MIDL. I encourage the authors to continue development and submit more complete work to future conferences.\n\nI thank the reviewers for also giving specific suggestions on how to improve the paper. ", "meta_review_readers": ["everyone"], "meta_review_writers": ["MIDL.io/2020/Conference/Program_Chairs", "MIDL.io/2020/Conference/Paper219/Area_Chairs"], "meta_review_reply_count": {"replyCount": 0}, "meta_review_url": ["https://openreview.net/forum?id=O6rt5JpTUl&noteId=4AYMNf6zrJz"], "decision": "reject"}