medXGAN: Visual Explanations for Medical Classifiers through a Generative Latent Space. Dravid, A., Schiffers, F., Gong, B., & Katsaggelos, A. K. In 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), volume 2022-June, pages 2935–2944, jun, 2022. IEEE. Paper doi abstract bibtex 1 download Despite the surge of deep learning in the past decade, some users are skeptical to deploy these models in practice due to their black-box nature. Specifically, in the medical space where there are severe potential repercussions, we need to develop methods to gain confidence in the models' decisions. To this end, we propose a novel medical imaging generative adversarial framework, medXGAN (medical eXplanation GAN), to visually explain what a medical classifier focuses on in its binary predictions. By encoding domain knowledge of medical images, we are able to disentangle anatomical structure and pathology, leading to fine-grained visualization through latent interpolation. Furthermore, we optimize the latent space such that interpolation explains how the features contribute to the classifier's output. Our method outperforms baselines such as Gradient-Weighted Class Activation Mapping (Grad-CAM) and Integrated Gradients in localization and explanatory ability. Additionally, a combination of the medXGAN with Integrated Gradients can yield explanations more robust to noise. The project page with code is available at: https://avdravid.github.io/medXGANpage/.
@inproceedings{dravid2022medxgan,
abstract = {Despite the surge of deep learning in the past decade, some users are skeptical to deploy these models in practice due to their black-box nature. Specifically, in the medical space where there are severe potential repercussions, we need to develop methods to gain confidence in the models' decisions. To this end, we propose a novel medical imaging generative adversarial framework, medXGAN (medical eXplanation GAN), to visually explain what a medical classifier focuses on in its binary predictions. By encoding domain knowledge of medical images, we are able to disentangle anatomical structure and pathology, leading to fine-grained visualization through latent interpolation. Furthermore, we optimize the latent space such that interpolation explains how the features contribute to the classifier's output. Our method outperforms baselines such as Gradient-Weighted Class Activation Mapping (Grad-CAM) and Integrated Gradients in localization and explanatory ability. Additionally, a combination of the medXGAN with Integrated Gradients can yield explanations more robust to noise. The project page with code is available at: https://avdravid.github.io/medXGANpage/.},
archivePrefix = {arXiv},
arxivId = {2204.05376},
author = {Dravid, Amil and Schiffers, Florian and Gong, Boqing and Katsaggelos, Aggelos K.},
booktitle = {2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},
doi = {10.1109/CVPRW56347.2022.00331},
eprint = {2204.05376},
isbn = {978-1-6654-8739-9},
issn = {21607516},
month = {jun},
pages = {2935--2944},
publisher = {IEEE},
title = {{medXGAN: Visual Explanations for Medical Classifiers through a Generative Latent Space}},
url = {https://ieeexplore.ieee.org/document/9857306/},
volume = {2022-June},
year = {2022}
}
Downloads: 1
{"_id":"uTmEBRNac8Jqe9W4r","bibbaseid":"dravid-schiffers-gong-katsaggelos-medxganvisualexplanationsformedicalclassifiersthroughagenerativelatentspace-2022","author_short":["Dravid, A.","Schiffers, F.","Gong, B.","Katsaggelos, A. K."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","abstract":"Despite the surge of deep learning in the past decade, some users are skeptical to deploy these models in practice due to their black-box nature. Specifically, in the medical space where there are severe potential repercussions, we need to develop methods to gain confidence in the models' decisions. To this end, we propose a novel medical imaging generative adversarial framework, medXGAN (medical eXplanation GAN), to visually explain what a medical classifier focuses on in its binary predictions. By encoding domain knowledge of medical images, we are able to disentangle anatomical structure and pathology, leading to fine-grained visualization through latent interpolation. Furthermore, we optimize the latent space such that interpolation explains how the features contribute to the classifier's output. Our method outperforms baselines such as Gradient-Weighted Class Activation Mapping (Grad-CAM) and Integrated Gradients in localization and explanatory ability. Additionally, a combination of the medXGAN with Integrated Gradients can yield explanations more robust to noise. The project page with code is available at: https://avdravid.github.io/medXGANpage/.","archiveprefix":"arXiv","arxivid":"2204.05376","author":[{"propositions":[],"lastnames":["Dravid"],"firstnames":["Amil"],"suffixes":[]},{"propositions":[],"lastnames":["Schiffers"],"firstnames":["Florian"],"suffixes":[]},{"propositions":[],"lastnames":["Gong"],"firstnames":["Boqing"],"suffixes":[]},{"propositions":[],"lastnames":["Katsaggelos"],"firstnames":["Aggelos","K."],"suffixes":[]}],"booktitle":"2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)","doi":"10.1109/CVPRW56347.2022.00331","eprint":"2204.05376","isbn":"978-1-6654-8739-9","issn":"21607516","month":"jun","pages":"2935–2944","publisher":"IEEE","title":"medXGAN: Visual Explanations for Medical Classifiers through a Generative Latent Space","url":"https://ieeexplore.ieee.org/document/9857306/","volume":"2022-June","year":"2022","bibtex":"@inproceedings{dravid2022medxgan,\nabstract = {Despite the surge of deep learning in the past decade, some users are skeptical to deploy these models in practice due to their black-box nature. Specifically, in the medical space where there are severe potential repercussions, we need to develop methods to gain confidence in the models' decisions. To this end, we propose a novel medical imaging generative adversarial framework, medXGAN (medical eXplanation GAN), to visually explain what a medical classifier focuses on in its binary predictions. By encoding domain knowledge of medical images, we are able to disentangle anatomical structure and pathology, leading to fine-grained visualization through latent interpolation. Furthermore, we optimize the latent space such that interpolation explains how the features contribute to the classifier's output. Our method outperforms baselines such as Gradient-Weighted Class Activation Mapping (Grad-CAM) and Integrated Gradients in localization and explanatory ability. Additionally, a combination of the medXGAN with Integrated Gradients can yield explanations more robust to noise. The project page with code is available at: https://avdravid.github.io/medXGANpage/.},\narchivePrefix = {arXiv},\narxivId = {2204.05376},\nauthor = {Dravid, Amil and Schiffers, Florian and Gong, Boqing and Katsaggelos, Aggelos K.},\nbooktitle = {2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)},\ndoi = {10.1109/CVPRW56347.2022.00331},\neprint = {2204.05376},\nisbn = {978-1-6654-8739-9},\nissn = {21607516},\nmonth = {jun},\npages = {2935--2944},\npublisher = {IEEE},\ntitle = {{medXGAN: Visual Explanations for Medical Classifiers through a Generative Latent Space}},\nurl = {https://ieeexplore.ieee.org/document/9857306/},\nvolume = {2022-June},\nyear = {2022}\n}\n","author_short":["Dravid, A.","Schiffers, F.","Gong, B.","Katsaggelos, A. K."],"key":"dravid2022medxgan","id":"dravid2022medxgan","bibbaseid":"dravid-schiffers-gong-katsaggelos-medxganvisualexplanationsformedicalclassifiersthroughagenerativelatentspace-2022","role":"author","urls":{"Paper":"https://ieeexplore.ieee.org/document/9857306/"},"metadata":{"authorlinks":{}},"downloads":1},"bibtype":"inproceedings","biburl":"https://sites.northwestern.edu/ivpl/files/2023/06/IVPL_Updated_publications-1.bib","dataSources":["KTWAakbPXLGfYseXn","ePKPjG8C6yvpk4mEK","ya2CyA73rpZseyrZ8","37Qfzv6wRptYkoSCL","zFPgsTDAW8aDnb5iN","E6Bth2QB5BYjBMZE7","nbnEjsN7MJhurAK9x","PNQZj6FjzoxxJk4Yi","7FpDWDGJ4KgpDiGfB","bod9ms4MQJHuJgPpp","QR9t5P2cLdJuzhfzK","D8k2SxfC5dKNRFgro","7Dwzbxq93HWrJEhT6","qhF8zxmGcJfvtdeAg","fvDEHD49E2ZRwE3fb","H7crv8NWhZup4d4by","DHqokWsryttGh7pJE","vRJd4wNg9HpoZSMHD","sYxQ6pxFgA59JRhxi","w2WahSbYrbcCKBDsC","XasdXLL99y5rygCmq","3gkSihZQRfAD2KBo3","t5XMbyZbtPBo4wBGS","bEpHM2CtrwW2qE8FP","teJzFLHexaz5AQW5z"],"keywords":[],"search_terms":["medxgan","visual","explanations","medical","classifiers","through","generative","latent","space","dravid","schiffers","gong","katsaggelos"],"title":"medXGAN: Visual Explanations for Medical Classifiers through a Generative Latent Space","year":2022,"downloads":1}