Classification of Liver Fibrosis From Heterogeneous Ultrasound Image

Abstract

With the advances in deep learning, including Convolutional Neural Networks (CNN), automated diagnosis technology using medical images has received considerable attention in medical science. In particular, in the field of ultrasound imaging, CNN trains the features of organs through an amount of image data, so that an expert-level automatic diagnosis is possible only with images of actual patients. However, CNN models are also trained on the features that reflect the inherent bias of the imaging machine used for image acquisition. In other words, when the domain of data used for training is different from that of data applied for an actual diagnosis, it is unclear whether consistent performance can be provided by the domain bias. Therefore, we investigate the effect of domain bias on the model with liver ultrasound imaging data obtained from multiple domains. We have constructed a dataset considering the manufacturer and the year of manufacturing of 8 ultrasound imaging machines. First, training and testing were performed by dividing the entire data, in a commonly used method. Second, we have utilized the training data constructed according to the number of domains for the machine learning process. Then we have measured and compared the performance on internal and external domain data. Through the above experiment, we have analyzed the effect of domains of data on model performance. We show that the performance scores evaluated with the internal domain data and the external domain data do not match. We especially show that the performance measured in the evaluation data including the internal domain was much higher than the performance measured in the evaluation data consisting of the external domain. We also show that 3-level classification performance is slightly improved over 5-level classification by mitigating class imbalance by integrating similar classes. The results highlight the need to develop a new methodology for mitigating the machine bias problem so that the model can work correctly even on external domain data, as opposed to the usual approach of constructing evaluation data in the same domain as the training data.INDEX TERMS Domain bias, multi-domain learning, ultrasonography, liver fibrosis.

TMM: A comprehensive CAD system for hepatic fibrosis 5‐grade METAVIR staging based on liver MRI

“…The performance of TMM for hepatic fibrosis staging is verified by comparing with existing methods including the classical methods for hepatic fibrosis CAD such as LDA, 11 KNN, 11 and SVM 10,11 with GLCM. In addition, there are some recent methods such as first‐order statistics, RL, and GLCM with MLP classifier 14 and ResNet50 17 . The same cross‐validation grouping is used in all comparison methods, and the parameters are adjusted to the best.…”

“…In addition, there are some recent methods such as firstorder statistics, RL, and GLCM with MLP classifier 14 and ResNet50. 17 The same cross-validation grouping is used in all comparison methods, and the parameters are adjusted to the best. From Table 6, TMM proposed by us obtains the highest accuracy for F1, F2, F3, F4, and the overall classification.…”

“…Balagourouchetty 16 proposed an ensemble FCNet classifier to classify hepatic lesions with the deep features extracted by GoogleNet-LReLU transfer learning. Joo et al 17 realized three-classification and five-classification of ultrasonic liver fibrosis images through CNN, and ResNet50 achieved the best results. Yu et al 18 built a classification model base on deep learning and scored the four stages of liver fibrosis (F0 vs. F1-4, F0-1 vs. F2-4, F0-2 vs. F3-4, and F0-3 vs. F4).…”

“…At the same time, most researches focus on binary classification, such as cirrhosis versus non-cirrhosis (F4 vs. F0-3), early versus advanced fibrosis (F0-2 vs. F3-4), substantial fibrosis or not (F2-4 vs. F0-1). In recent years, there are few studies 17,20,21 focusing on multiple classification of hepatic fibrosis, and an accurate comprehensive five-classification with liver MRIs is expected for precision diagnosis. However, hepatic fibrosis staging is challenging as the different fibrosis stages are very similar 12 and sensitive radiomics features extraction and classification are needed.…”

“…Balagourouchetty 16 proposed an ensemble FCNet classifier to classify hepatic lesions with the deep features extracted by GoogleNet‐LReLU transfer learning. Joo et al 17 . realized three‐classification and five‐classification of ultrasonic liver fibrosis images through CNN, and ResNet50 achieved the best results.…”