Objective: Accurate segmentation of liver tumors, which could help physicians make appropriate treatment decisions and assess the effectiveness of surgical treatment, is crucial for the clinical diagnosis of liver cancer. In this study, we propose a 4-dimensional (4D) deep learning model based on 3D convolution and convolutional long short-term memory (C-LSTM) for hepatocellular carcinoma (HCC) lesion segmentation. Methods: The proposed deep learning model utilizes 4D information on dynamic contrast enhanced (DCE) magnetic resonance imaging (MRI) images to assist liver tumor segmentation. Specifically, a shallow U-net based 3D CNN module was designed to extract 3D spatial domain features from each DCE phase, followed by a 4-layer C-LSTM network module for time domain information exploitation. The combined information of multi-phase DCE images and the manner by which tissue imaging features change on multi-contrast images allow the network to more effectively learn the characteristics of HCC, resulting in better segmentation performance. Results: The proposed model achieved a Dice score of 0.825± 0.077, a Hausdorff distance of 12.84± 8.14 mm, and a volume similarity of 0.891± 0.080 for liver tumor segmentation, which outperformed the 3D U-net model, RA-UNet model and other models in the ablation study in both internal and external test sets. Moreover, the performance of the proposed model is comparable to the nnU-Net model, which showed state-of-the-art performance in many segmentation tasks, with significantly reduced prediction time. Conclusion: The proposed 3D convolution and C-LSTM based model can achieve accurate segmentation of HCC lesions.

Deep learning based generative adversarial networks (GAN) can effectively perform image reconstruction with under-sampled MR data. In general, a large number of training samples are required to improve the reconstruction performance of a certain model. However, in real clinical applications, it is difficult to obtain tens of thousands of raw patient data to train the model since saving k-space data is not in the routine clinical flow. Therefore, enhancing the generalizability of a network based on small samples is urgently needed. In this study, three novel applications were explored based on parallel imaging combined with the GAN model (PI-GAN) and transfer learning. The model was pre-trained with public Calgary brain images and then fine-tuned for use in (1) patients with tumors in our center; (2) different anatomies, including knee and liver; (3) different k-space sampling masks with acceleration factors (AFs) of 2 and 6. As for the brain tumor dataset, the transfer learning results could remove the artifacts found in PI-GAN and yield smoother brain edges. The transfer learning results for the knee and liver were superior to those of the PI-GAN model trained with its own dataset using a smaller number of training cases. However, the learning procedure converged more slowly in the knee datasets compared to the learning in the brain tumor datasets. The reconstruction performance was improved by transfer learning both in the models with AFs of 2 and 6. Of these two models, the one with AF = 2 showed better results. The results also showed that transfer learning with the pre-trained model could solve the problem of inconsistency between the training and test datasets and facilitate generalization to unseen data.

In this study, we proposed a model combing parallel imaging (PI) with generative adversarial network (GAN) architecture (PIC-GAN) for accelerated multi-channel magnetic resonance imaging (MRI) reconstruction. This model integrated data fidelity and regularization terms into the generator to benefit from multi-coils information and provide an “end-to-end” reconstruction. Besides, to better preserve image details during reconstruction, we combined the adversarial loss with pixel-wise loss in both image and frequency domains. The proposed PIC-GAN framework was evaluated on abdominal and knee MRI images using 2, 4 and 6-fold accelerations with different undersampling patterns. The performance of the PIC-GAN was compared to the sparsity-based parallel imaging (L1-ESPIRiT), the variational network (VN), and conventional GAN with single-channel images as input (zero-filled (ZF)-GAN). Experimental results show that our PIC-GAN can effectively reconstruct multi-channel MR images at a low noise level and improved structure similarity of the reconstructed images. PIC-GAN has yielded the lowest Normalized Mean Square Error (in $\times {10}^{-5}$) (PIC-GAN: 0.58 ± 0.37, ZF-GAN: 1.93 ± 1.41, VN: 1.87 ± 1.28, L1-ESPIRiT: 2.49 ± 1.04 for abdominal MRI data and PIC-GAN: 0.80 ± 0.26, ZF-GAN: 0.93 ± 0.29, VN:1.18 ± 0.31, L1-ESPIRiT: 1.28 ± 0.24 for knee MRI data) and the highest Peak Signal to Noise Ratio (PIC-GAN: 34.43 ± 1.92, ZF-GAN: 31.45 ± 4.0, VN: 29.26 ± 2.98, L1-ESPIRiT: 25.40 ± 1.88 for abdominal MRI data and PIC-GAN: 34.10 ± 1.09, ZF-GAN: 31.47 ± 1.05, VN: 30.01 ± 1.01, L1-ESPIRiT: 28.01 ± 0.98 for knee MRI data) compared to ZF-GAN, VN and L1-ESPIRiT with an under-sampling factor of 6. The proposed PIC-GAN framework has shown superior reconstruction performance in terms of reducing aliasing artifacts and restoring tissue structures as compared to other conventional and state-of-the-art reconstruction methods.

This study aimed to explore the added value of viscoelasticity measured by magnetic resonance elastography (MRE) in the prediction of Ki-67 expression in hepatocellular carcinoma (HCC) using a deep learning combined radiomics (DLCR) model. This retrospective study included 108 histopathology-proven HCC patients (93 males; age, 59.6 ± 11.0 years) who underwent preoperative MRI and MR elastography. They were divided into training (n = 87; 61.0 ± 9.8 years) and testing (n = 21; 60.6 ± 10.1 years) cohorts. An independent validation cohort including 43 patients (60.1 ± 11.3 years) was included for testing. A DLCR model was proposed to predict the expression of Ki-67 with cMRI, including T2W, DW, and dynamic contrast enhancement (DCE) images as inputs. The images of the shear wave speed (c-map) and phase angle (φ-map) derived from MRE were also fed into the DLCR model. The Ki-67 expression was classified into low and high groups with a threshold of 20%. Both c and φ values were ranked within the top six features for Ki-67 prediction with random forest selection, which revealed the value of MRE-based viscosity for the assessment of tumor proliferation status in HCC. When comparing the six CNN models, Xception showed the best performance for classifying the Ki-67 expression, with an AUC of 0.80 ± 0.03 (CI: 0.79–0.81) and accuracy of 0.77 ± 0.04 (CI: 0.76–0.78) when cMRI were fed into the model. The model with all modalities (MRE, AFP, and cMRI) as inputs achieved the highest AUC of 0.90 ± 0.03 (CI: 0.89–0.91) in the validation cohort. The same finding was observed in the independent testing cohort, with an AUC of 0.83 ± 0.03 (CI: 0.82–0.84). The shear wave speed and phase angle improved the performance of the DLCR model significantly for Ki-67 prediction, suggesting that MRE-based c and φ-maps can serve as important parameters to assess the tumor proliferation status in HCC.