ABSTRACT
BACKGROUND AND OBJECTIVE: Hepatocellular carcinoma is a common disease with high mortality. Through deep learning methods to analyze HCC CT, the screening classification and prognosis model of HCC can be established, which further promotes the development of computer-aided diagnosis and treatment in the treatment of HCC. However, there are significant challenges in the actual establishment of HCC auxiliary diagnosis model due to data imbalance, especially for rare subtypes of HCC and underrepresented demographic groups. This study proposes a GAN model aimed at overcoming these obstacles and improving the accuracy of HCC auxiliary diagnosis. METHODS: In order to generate liver and tumor images close to the real distribution. Firstly, we construct a new gradient transfer sampling module to improve the lack of texture details and excessive gradient transfer parameters of the deep model; Secondly, we construct an attention module with spatial and cross-channel feature extraction ability to improve the discriminator's ability to distinguish images; Finally, we design a new loss function for liver tumor imaging features to constrain the model to approach the real tumor features in iterations. RESULTS: In qualitative analysis, the images synthetic by our method closely resemble the real images in liver parenchyma, blood vessels, tumors, and other parts. In quantitative analysis, the optimal results of FID, PSNR, and SSIM are 75.73, 22.77, and 0.74, respectively. Furthermore, our experiments establish classification models for imbalanced data and enhanced data, resulting in an increase in accuracy rate by 21%-34%, an increase in AUC by 0.29 - 0.33, and an increase in specificity to 0.89. CONCLUSION: Our solution provides a variety of training data sources with low cost and high efficiency for the establishment of classification or prognostic models for imbalanced data.
ABSTRACT
BACKGROUND: The macrotrabecular-massive (MTM) is a special subtype of hepatocellular carcinoma (HCC), which has commonly a dismal prognosis. This study aimed to develop a multitask deep learning radiomics (MDLR) model for predicting MTM and HCC patients' prognosis after hepatic arterial infusion chemotherapy (HAIC). METHODS: From June 2018 to March 2020, 158 eligible patients with HCC who underwent surgery were retrospectively enrolled in MTM related cohorts, and 752 HCC patients who underwent HAIC were included in HAIC related cohorts during the same period. DLR features were extracted from dual-phase (arterial phase and venous phase) contrast-enhanced computed tomography (CECT) of the entire liver region. Then, an MDLR model was used for the simultaneous prediction of the MTM subtype and patient prognosis after HAIC. The MDLR model for prognostic risk stratification incorporated DLR signatures, clinical variables and MTM subtype. FINDINGS: The predictive performance of the DLR model for the MTM subtype was 0.968 in the training cohort [TC], 0.912 in the internal test cohort [ITC] and 0.773 in the external test cohort [ETC], respectively. Multivariable analysis identified portal vein tumor thrombus (PVTT) (p = 0.012), HAIC response (p < 0.001), HAIC sessions (p < 0.001) and MTM subtype (p < 0.001) as indicators of poor prognosis. After incorporating DLR signatures, the MDLR model yielded the best performance among all models (AUC, 0.855 in the TC, 0.805 in the ITC and 0.792 in the ETC). With these variables, the MDLR model provided two risk strata for overall survival (OS) in the TC: low risk (5-year OS, 44.9%) and high risk (5-year OS, 4.9%). INTERPRETATION: A tool based on MDLR was developed to consider that the MTM is an important prognosis factor for HCC patients. MDLR showed outstanding performance for the prognostic risk stratification of HCC patients who underwent HAIC and may help physicians with therapeutic decision making and surveillance strategy selection in clinical practice.
