B-mode ultrasound based CAD for liver cancers via multi-view privileged information learning.

B-mode ultrasound-based computer-aided diagnosis model can help sonologists improve the diagnostic performance for liver cancers, but it generally suffers from the bottleneck due to the limited structure and internal echogenicity information in B-mode ultrasound images. Contrast-enhanced ultrasound images provide additional diagnostic information on dynamic blood perfusion of liver lesions for B-mode ultrasound images with improved diagnostic accuracy. Since transfer learning has indicated its effectiveness in promoting the performance of target computer-aided diagnosis model by transferring knowledge from related imaging modalities, a multi-view privileged information learning framework is proposed to improve the diagnostic accuracy of the single-modal B-mode ultrasound-based diagnosis for liver cancers. This framework can make full use of the shared label information between the paired B-mode ultrasound images and contrast-enhanced ultrasound images to guide knowledge transfer It consists of a novel supervised dual-view deep Boltzmann machine and a new deep multi-view SVM algorithm. The former is developed to implement knowledge transfer from the multi-phase contrast-enhanced ultrasound images to the B-mode ultrasound-based diagnosis model via a feature-level learning using privileged information paradigm, which is totally different from the existing learning using privileged information paradigm that performs knowledge transfer in the classifier. The latter further fuses and enhances feature representation learned from three pre-trained supervised dual-view deep Boltzmann machine networks for the classification task. An experiment is conducted on a bimodal ultrasound liver cancer dataset. The experimental results show that the proposed framework outperforms all the compared algorithms with the best classification accuracy of 88.91 ± 1.52%, sensitivity of 88.31 ± 2.02%, and specificity of 89.50 ± 3.12%. It suggests the effectiveness of our proposed MPIL framework for the BUS-based CAD of liver cancers.

Credit rating in dynamic response to the nature of firms and the business model of rating agencies: evidence from the Chinese bond market.

This paper takes the A-share listed companies that issued credit bonds from 2010 to 2021 as the sample to test the probability and degree of credit rating change throughout the enterprise life cycle using the ordered logit and breakpoint regression models. Further, we study the heterogeneity of the above performance from payment models and firm natures. The results show that the credit rating inflation problem generally exists in all stages of the enterprise life cycle. The inflation is lower in the investor-pays model (state-owned enterprises), while the opposite results occur for the issuer-pays model (non-state-owned enterprises). Specifically, (1) the probability of a higher credit rating and the increased credit ratings show as an 'inverse U' in the enterprise life cycle. Credit rating increases if the enterprise successfully enters the growth phase, decreases if the enterprise fell into the decline phase. (2) In the investor-pays model, enterprises have a greater probability of obtaining a higher credit rating in the mature phase and a lower credit rating during the decline period. In the issuer-pays model, although the enterprise gets a smaller credit rating due to falling into the decline phase, the credit rating still has a high probability of belonging to a high credit rating. (3) State-owned enterprises have a higher probability of obtaining a high credit rating in the mature period and are more likely to have a low credit rating in the decline period. Generally, their credit rating quality is better than that of non-state-owned enterprises. In addition, in the context of the financing pressure period, the credit rating of non-state-owned enterprises decreases as they drop into the decline phase.

Diagnosis of Infantile Hip Dysplasia With B-Mode Ultrasound via Two-Stage Meta-Learning Based Deep Exclusivity Regularized Machine.

The B-mode ultrasound (BUS) based computer-aided diagnosis (CAD) has shown its effectiveness for developmental dysplasia of the hip (DDH) in infants. In this work, a two-stage meta-learning based deep exclusivity regularized machine (TML-DERM) is proposed for the BUS-based CAD of DDH. TML-DERM integrates deep neural network (DNN) and exclusivity regularized machine into a unified framework to simultaneously improve the feature representation and classification performance. Moreover, the first-stage meta-learning is mainly conducted on the DNN module to alleviate the overfitting issue caused by the significantly increased parameters in DNN, and a random sampling strategy is adopted to self-generate the meta-tasks; while the second-stage meta-learning mainly learns the combination of multiple weak classifiers by a weight vector to improve the classification performance, and also optimizes the unified framework again. The experimental results on a DDH ultrasound dataset show the proposed TML-DERM algorithm achieves the superior classification performance with the mean accuracy of 85.89%, sensitivity of 86.54%, and specificity of 85.23%.

B-Mode Ultrasound Based Diagnosis of Liver Cancer With CEUS Images as Privileged Information.

Contrast-enhanced ultrasound (CEUS) is a valuable imaging modality for diagnosis of liver cancers. However, the complexity of CEUS-based diagnosis limits its wide application, and the B-mode ultrasound (BUS) is still the most popular diagnosis modality in clinical practice. In order to promote BUS-based computer-aided diagnosis (CAD) for liver cancers, we propose a learning using privileged information (LUPI) based CAD with BUS as the diagnosis modality and CEUS as PI. Particularly, the multimodal restricted Boltzmann machine (MRBM) works as a LUPI paradigm. That is, one BUS image and three CEUS images from the arterial phase, portal venous phase and delayed phase, respectively, are used to train three multimodal restricted Boltzmann machine (MRBM) models during training stage, but only the BUS data will be fed to MRBM to generate new feature representation at testing phase. A multiple empirical kernel learning machine (MEKLM) classifier is then performed on three new feature vectors from three MRBM models for classification of liver cancers. The experimental results show that the proposed MRBM-MEKLM algorithm outperforms all the compared algorithms, suggesting the effectiveness of the proposed LUPI-based CAD for liver cancer.

Multiple Empirical Kernel Mapping Based Broad Learning System for Classification of Parkinson's Disease With Transcranial Sonography.

Transcranial sonography (TCS) has become more popular for diagnosis of Parkinson's disease (PD), and the TCS-based computer-aided diagnosis (CAD) for PD also attracts considerable attention, in which classifier is a critical component. Broad learning system (BLS) is a newly proposed single layer feedforward neural network for classification. In BLS, the original input features are mapped to several new feature representations to form the feature nodes, and then these mapped features are expanded to enhancement nodes by random mapping in a wide sense. However, random mapping performed for enhancement nodes is too simple and the generated features lack interpretability together with relative low representation. In this work, we propose a multiple empirical kernel mapping (MEKM) based BLS (MEKM-BLS) algorithm, which adopts MEKM to map the data of feature nodes to enhancement nodes. MEKM-BLS then has more meaningful enhancement layer in feedforward neural network. Moreover, the experiment for PD diagnosis with TCS shows that MEKM-BLS achieves superior performance to the original BLS algorithm.