Improving B-mode ultrasound diagnostic performance for focal liver lesions using deep learning: A multicentre study

Abstract

Background: The diagnosis performance of B-mode ultrasound (US) for focal liver lesions (FLLs) is relatively limited. We aimed to develop a deep convolutional neural network of US (DCNN-US) for aiding radiologists in classification of malignant from benign FLLs. Materials and methods: This study was conducted in 13 hospitals and finally 2143 patients with 24,343 US images were enrolled. Patients who had non-cystic FLLs with pathological results were enrolled. The FLLs from 11 hospitals were randomly divided into training and internal validations (IV) cohorts with a 4:1 ratio for developing and evaluating DCNN-US. Diagnostic performance of the model was verified using external validation (EV) cohort from another two hospitals. The diagnosis value of DCNN-US was compared with that of contrast enhanced computed tomography (CT)/magnetic resonance image (MRI) and 236 radiologists, respectively. Findings: The AUC of Model LBC for FLLs was 0.924 (95% CI: 0.889À0.959) in the EV cohort. The diagnostic sensitivity and specificity of Model LBC were superior to 15-year skilled radiologists (86.5% vs 76.1%, p = 0.0084 and 85.5% vs 76.9%, p = 0.0051, respectively). Accuracy of Model LBC was comparable to that of contrast enhanced CT (both 84.7%) but inferior to contrast enhanced MRI (87.9%) for lesions detected by US. Interpretation: DCNN-US with high sensitivity and specificity in diagnosing FLLs shows its potential to assist less-experienced radiologists in improving their performance and lowering their dependence on sectional imaging in liver cancer diagnosis.

US-Based Deep Learning Model for Differentiating Hepatocellular Carcinoma (HCC) From Other Malignancy in Cirrhotic Patients

“…Two experienced radiologists reviewed all US images, who were not participated in the image acquisition and blinded to clinical information and final diagnoses of each patient. The review for US features was in concordance with a prior study (14). In brief, echogenicity was defined as mixed, hypoechoic or hyperechoic when comparing with the echogenicity of surrounding parenchyma.…”

“…The biggest challenge in traditional machine learning and texture analysis is operator-dependently predefining region-of-interest (ROI), which could be overcome by free-hand data mining and navigating system of DLM ( 23 ). Previous literatures reported great potential of deep-learning information in differentiating FLLs based on CECT ( 24 ), CEMRI ( 25 ) and US modality ( 14 ). However, prior research groups failed to especially focus on the role of DLM in cirrhotic patients, who may have different liver cancer dynamics from normal ones ( 26 ).…”

“…However, prior research groups failed to especially focus on the role of DLM in cirrhotic patients, who may have different liver cancer dynamics from normal ones ( 26 ). Meanwhile, when lesions were concurrent with cirrhosis, US had variable and suboptimal value in diagnosing HCCs with high confidence ( 14 ). Thus DLM may exist as a helpful tool to dig much more data hidden in US images than radiologists could observe.…”

AI-Based Radiological Imaging for HCC: Current Status and Future of Ultrasound

“…A recent retrospective study analyzed US images of 668 patients with primary liver cancer, consisting of 531 HCC patients, 48 cHCC-CC patients, and 89 CC patients, and found that the overall performance of the radiomics model in identifying different histopathological types of primary liver cancer yielded AUCs of 0.854 (training cohort) and 0.775 (test cohort) in the HCC vs. non-HCC radiomics model, and 0.920 (training cohort) and 0.728 (test cohort) in the CC vs. cHCC-CC radiomics model [ 79 ]. Furthermore, a multicenter study performed in 13 hospitals, including 2143 patients (24343 sonograms), has shown that deep convolutional neural network of US may have the potential to assist less experienced radiologists in improving their performance and lowering their dependence on sectional imaging in liver cancer diagnosis [ 47 ].…”

Deep learning promotes B-mode ultrasound screening for focal liver lesions

“…In this article of EBioMedicine, Yang and colleagues report the excellent diagnostic performance of the developed deep convolutional neural network of US (DCNN-US) in classification of malignant from benign FLLs using 11 standard still US images and clinical-ultrasonic factors [5]. It is noteworthy that DCNN-US showed higher diagnostic performance compared to experienced radiologists and comparable diagnostic performance to contrast-enhanced CT for lesions detected by US in a large external validation cohort obtained from a prospective multicentre study.…”