Cheng Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Cheng Li has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Jiangli Lin
College of Biomedical Engineering, Sichuan University, Chengdu, China
2 shared publications - Ke Chen
College of Biomedical Engineering, Sichuan University, Chengdu 610065, China
2 shared publications - Zhan Hua
Department of General Surgery, China-Japan Friendship Hospital, Beijing, 100029, People's Republic of China
2 shared publications - Heqing Zhang
Department of Ultrasound, West China Hospital, Sichuan University, Chengdu 610065, China
1 shared publication - Jie Luo
College of Biomedical Engineering, Sichuan University, Chengdu 610065, China
1 shared publication - Lin Han
College of Biomedical Engineering, Sichuan University, Chengdu 610065, China
1 shared publication
Latest Publications
- Open Access
2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
Deep learning (DL) models in breast ultrasound (BUS) image analysis face challenges with data imbalance and limited atypical tumor samples. Generative Adversarial Networks (GAN) address these challenges by providing efficient data augmentation for small datasets. However, current GAN approaches fail to capture the structural features of BUS and generated images lack structural legitimacy and are unrealistic. Furthermore, generated images require manual annotation for different downstream tasks before they can be used. Therefore, we propose a two-stage GAN framework, 2s-BUSGAN, for generating annotated BUS images. It consists of the Mask Generation Stage (MGS) and the Image Generation Stage (IGS), generating benign and malignant BUS images using corresponding tumor contours. Moreover, we employ a Feature-Matching Loss (FML) to enhance the quality of generated images and utilize a Differential Augmentation Module (DAM) to improve GAN performance on small datasets. We conduct experiments on two datasets, BUSI and Collected. Moreover, results indicate that the quality of generated images is improved compared with traditional GAN methods. Additionally, our generated images underwent evaluation by ultrasound experts, demonstrating the possibility of deceiving doctors. A comparative evaluation showed that our method also outperforms traditional GAN methods when applied to training segmentation and classification models. Our method achieved a classification accuracy of 69% and 85.7% on two datasets, respectively, which is about 3% and 2% higher than that of the traditional augmentation model. The segmentation model trained using the 2s-BUSGAN augmented datasets achieved DICE scores of 75% and 73% on the two datasets, respectively, which were higher than the traditional augmentation methods. Our research tackles imbalanced and limited BUS image data challenges. Our 2s-BUSGAN augmentation method holds potential for enhancing deep learning model performance in the field.
Open Access2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets
Deep learning (DL) models in breast ultrasound (BUS) image analysis face challenges with data imbalance and limited atypical tumor samples. Generative Adversarial Networks (GAN) address these challenges by providing efficient data augmentation for small datasets. However, current GAN approaches fail to capture the structural features of BUS and generated images lack structural legitimacy and are unrealistic. Furthermore, generated images require manual annotation for different downstream tasks before they can be used. Therefore, we propose a two-stage GAN framework, 2s-BUSGAN, for generating annotated BUS images. It consists of the Mask Generation Stage (MGS) and the Image Generation Stage (IGS), generating benign and malignant BUS images using corresponding tumor contours. Moreover, we employ a Feature-Matching Loss (FML) to enhance the quality of generated images and utilize a Differential Augmentation Module (DAM) to improve GAN performance on small datasets. We conduct experiments on two datasets, BUSI and Collected. Moreover, results indicate that the quality of generated images is improved compared with traditional GAN methods. Additionally, our generated images underwent evaluation by ultrasound experts, demonstrating the possibility of deceiving doctors. A comparative evaluation showed that our method also outperforms traditional GAN methods when applied to training segmentation and classification models. Our method achieved a classification accuracy of 69% and 85.7% on two datasets, respectively, which is about 3% and 2% higher than that of the traditional augmentation model. The segmentation model trained using the 2s-BUSGAN augmented datasets achieved DICE scores of 75% and 73% on the two datasets, respectively, which were higher than the traditional augmentation methods. Our research tackles imbalanced and limited BUS image data challenges. Our 2s-BUSGAN augmentation method holds potential for enhancing deep learning model performance in the field.
- Open Access
- 18 June 2018
- Published by Springer Nature in BioMedical Engineering OnLine
- Vol. 17 (1) , 1-17
A novel TIRADS of US classification
Thyroid imaging reporting and data system (TIRADS) is the assessment of a risk stratification of thyroid nodules, usually using a score. However, there is no consensus as to the version of TIRADS for reporting the results of thyroid ultrasound in clinic. The objective of this study is to develop a practical TIRADS with which to categorize thyroid nodules and stratify their malignant risk. A TIRADS scoring system was developed to provide more decision levels than standard scoring through the selection of the ultrasound features which include the calcification shape, margins, taller-than-wide, internal echo, blood flow quantization of features, setting of the weight, and calculation of the score. Ultimately, the accuracy of our TIRADS was evaluated by comparing with the results of current vision of TIRADS and thyroid radiologist in 153 patients who had US-guided fine-needle aspiration biopsy. Classification results showed that the total accuracy reached 97% (100% of malignant and 95% of the benign) in 153 cases (benign:78, malignant:75). The percentages of malignancy is defined in our TIRADS were as follows: TIRADS 2 (0% malignancy), TIRADS 3 (3.6% malignancy), TIRADS 4 (17–75% malignancy), and TIRADS 5 (98% malignancy). We established a novel TIRADS to predict the malignancy risk of the thyroid nodules based on six categories US features by a scoring system, which included a standardized vocabulary and score and a quantified risk assessment. The results showed that objective quantitative classification of thyroid nodules by our TIRADS can be useful in guiding management decisions.
- 18 June 2018
- Published by Springer Nature in BioMedical Engineering OnLine
- Vol. 17 (1) , 1-17
Open AccessA novel TIRADS of US classification
Thyroid imaging reporting and data system (TIRADS) is the assessment of a risk stratification of thyroid nodules, usually using a score. However, there is no consensus as to the version of TIRADS for reporting the results of thyroid ultrasound in clinic. The objective of this study is to develop a practical TIRADS with which to categorize thyroid nodules and stratify their malignant risk. A TIRADS scoring system was developed to provide more decision levels than standard scoring through the selection of the ultrasound features which include the calcification shape, margins, taller-than-wide, internal echo, blood flow quantization of features, setting of the weight, and calculation of the score. Ultimately, the accuracy of our TIRADS was evaluated by comparing with the results of current vision of TIRADS and thyroid radiologist in 153 patients who had US-guided fine-needle aspiration biopsy. Classification results showed that the total accuracy reached 97% (100% of malignant and 95% of the benign) in 153 cases (benign:78, malignant:75). The percentages of malignancy is defined in our TIRADS were as follows: TIRADS 2 (0% malignancy), TIRADS 3 (3.6% malignancy), TIRADS 4 (17–75% malignancy), and TIRADS 5 (98% malignancy). We established a novel TIRADS to predict the malignancy risk of the thyroid nodules based on six categories US features by a scoring system, which included a standardized vocabulary and score and a quantified risk assessment. The results showed that objective quantitative classification of thyroid nodules by our TIRADS can be useful in guiding management decisions.