What And How Other Datasets Can Be Leveraged For Medical Imaging Classification

Shang, Hong; Sun, Zhongqian; Fu, Xinghui; Zhang, Zijian; Yang, Wei

doi:10.1109/isbi.2019.8759148

Cited by 3 publications

(6 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The network is then transferred as a feature extractor on the target task (tumour classification). Shang et al [17] use several datasets (including some unrelated to their target task such as Dogs vs. cats) and compare ImageNet and domain-specific pre-training in order to tackle colonoscopy image classification. They also show that pre-training on domain-specific data yield superior performance compared to using ImageNet.…”

Section: Related Workmentioning

confidence: 99%

“…The structure of our multi-task neural network is similar to those of [17] and [34] and is guided by the objective of pretraining a network for transfer. Therefore, we have adopted the architecture presented in Figure 1.…”

Section: A Multi-task Architecturementioning

confidence: 99%

“…As the source task must be a large dataset, the most common choice is using ImageNet, a classification dataset containing more than 1 million natural images organized into 1000 classes, as a source. Although ImageNet images are very dissimilar to digital pathology images, it has been shown that transferring from the former can still boost performance of deep learning methods on the latter [9], [17].…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, this question arises: could we get even better performance from transfer learning by using digital pathology pre-trained models instead of an ImageNet one ? Some works, e.g., [17], [19]- [21], have advanced that domainspecific pre-training is indeed beneficial but, to the best of our knowledge, there is no in-depth study that attempted to answer this question in digital pathology. The main obstacle preventing this question to be answered is the lack of a large and versatile dataset like ImageNet.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Multi-Task Pre-Training of Deep Neural Networks for Digital Pathology

Mormont

Geurts

Marée

2021

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

In this work, we investigate multi-task learning as a way of pre-training models for classification tasks in digital pathology. It is motivated by the fact that many small and medium-size datasets have been released by the community over the years whereas there is no large scale dataset similar to ImageNet in the domain. We first assemble and transform many digital pathology datasets into a pool of 22 classification tasks and almost 900k images. Then, we propose a simple architecture and training scheme for creating a transferable model and a robust evaluation and selection protocol in order to evaluate our method. Depending on the target task, we show that our models used as feature extractors either improve significantly over ImageNet pre-trained models or provide comparable performance. Finetuning improves performance over feature extraction and is able to recover the lack of specificity of ImageNet features, as both pre-training sources yield comparable performance.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: A Multi-task Architecturementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Multi-Task Pre-Training of Deep Neural Networks for Digital Pathology

Mormont

Geurts

Marée

2021

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

show abstract

“…Shang et al trained multiple 121-layer DenseNet models [129] with different combinations of five training datasets (NBI Colonoscopy, white-light Colonoscopy, Esophagogastroduodenoscopy, Skin Lesion, and ImageNet) [130]. The test dataset defines non-adenomatous polyp images as benign and adenomatous polyps and cancer images as malignant.…”

Section: Detection Of Other Types Of Abnormalitymentioning

confidence: 99%

Artificial Intelligence for Colonoscopy: Past, Present, and Future

Tavanapong

Riegler

et al. 2022

IEEE J. Biomed. Health Inform.

View full text Add to dashboard Cite

During the past decades, many automated image analysis methods have been developed for colonoscopy. Realtime implementation of the most promising methods during colonoscopy has been tested in clinical trials, including several recent multi-center studies. All trials have shown results that may contribute to prevention of colorectal cancer. We summarize the past and present development of colonoscopy video analysis methods, focusing on two categories of artificial intelligence (AI) technologies used in clinical trials. These are (1) analysis and feedback for improving colonoscopy quality and (2) detection of abnormalities. Our survey includes methods that use traditional machine learning algorithms on carefully designed hand-crafted features as well as recent deep-learning methods. Lastly, we present the gap between current state-of-the-art technology and desirable clinical features and conclude with future directions of endoscopic AI technology development that will bridge the current gap.

show abstract