Artificial intelligence (AI) models for the ultrasonographic diagnosis of liver tumors and comparison of diagnostic accuracies between AI and human experts

Abstract

Background Ultrasonography (US) is widely used for the diagnosis of liver tumors. However, the accuracy of the diagnosis largely depends on the visual perception of humans. Hence, we aimed to construct artificial intelligence (AI) models for the diagnosis of liver tumors in US. Methods We constructed three AI models based on still B-mode images: model-1 using 24,675 images, model-2 using 57,145 images, and model-3 using 70,950 images. A convolutional neural network was used to train the US images. The four-class liver tumor discrimination by AI, namely, cysts, hemangiomas, hepatocellular carcinoma, and metastatic tumors, was examined. The accuracy of the AI diagnosis was evaluated using tenfold cross-validation. The diagnostic performances of the AI models and human experts were also compared using an independent test cohort of video images. Results The diagnostic accuracies of model-1, model-2, and model-3 in the four tumor types are 86.8%, 91.0%, and 91.1%, whereas those for malignant tumor are 91.3%, 94.3%, and 94.3%, respectively. In the independent comparison of the AIs and physicians, the percentages of correct diagnoses (accuracies) by the AIs are 80.0%, 81.8%, and 89.1% in model-1, model-2, and model-3, respectively. Meanwhile, the median percentages of correct diagnoses are 67.3% (range 63.6%–69.1%) and 47.3% (45.5%–47.3%) by human experts and non-experts, respectively. Conclusion The performance of the AI models surpassed that of human experts in the four-class discrimination and benign and malignant discrimination of liver tumors. Thus, the AI models can help prevent human errors in US diagnosis.

Deep Learning for Approaching Hepatocellular Carcinoma Ultrasound Screening Dilemma: Identification of α-Fetoprotein-Negative Hepatocellular Carcinoma From Focal Liver Lesion Found in High-Risk Patients

“…With help of a huge sample of FLLs, the model based on the DL method in their study is able to do multi-grouping tasks, and diagnosis performance for every subtype was achievable. The overall accuracy of 89.1% for four discrimination was achieved (33).…”

“…In the testing cohort, both studies showed moderate AUC for each model (0.728-0.775), considering that as each diagnosis process goes through two to three tandem models, the accuracy for differential diagnosis of subtypes might not be ideal. A multicenter study used over 150,000 images focused on the differentiation of FLL subtypes, including cyst, hemangioma, HCC, and liver metastasis (33). With help of a huge sample of FLLs, the model based on the DL method in their study is able to do multi-grouping tasks, and diagnosis performance for every subtype was achievable.…”

Clinical Applications of Artificial Intelligence in Gastroenterology: Excitement and Evidence

“…Until now, most studies had focused on the liver, with AI improving the accuracy of an experienced observer, classifying benign from malignant solid liver lesions. 17,18 In addition, DL radiomics of elastography data showed similar diagnostic efficacy with liver biopsy for assessing cirrhosis and advanced fibrosis, 19 potentially eliminating the need for invasive testing.…”

Artificial intelligence models for the diagnosis and management of liver diseases

“…83.6%). In addition, the estimation probability of AI for correct diagnosis increases with an increase in the amount of training data, indicating that a highly reliable estimation can be acquired after an increase in training data [51].…”

“…A variety of AI models for diagnosis of liver disease have been reported; some of them reportedly outperform human experts based on their performance. However, the value of the output from AI could be more informative for beginners and non-experts because AI generally shows much higher performance than beginners and non-experts, compared to the experts, especially in the field of imaging diagnosis [42,51]. Another study indicate that the AI for diagnosing focal liver lesions in B-mode US shows the potential to assist less-experienced radiologists in improving their performance and lowering their dependence on sectional imaging in liver cancer diagnosis [41], although the final responsibility for medical decision is on the medical staff.…”