Objectives Noninvasive fetal electrocardiography (FECG) offers many advantages over alternative fetal monitoring techniques in evaluating fetal health conditions. However, it is difficult to extract a clean FECG signal with morphological features from an abdominal ECG recorded at the maternal abdomen; the signal is usually contaminated by the maternal ECG and various noises. The aim of the work is to extract an FECG signal that preserves the morphological features from the mother’s abdominal ECG recording, which allows for accurately estimating the fetal heart rate (FHR) and analyzing the waveforms of the fetal ECG. Methods We propose a novel nonlinear adaptive noise cancelling framework (ANC) based on a temporal convolutional neural network (CNN) to effectively extract fetal ECG signals from mothers’ abdominal ECG recordings. The proposed framework consists of a two-stage network, using the ANC architecture; one network is for the maternal ECG component elimination and the other is for the residual noise component removal of the extracted fetal ECG signal. Then, JADE (one of the blind source separation algorithms) is applied as a postprocessing step to produce a clean fetal ECG signal. Results Synthetic ECG data (FECGSYNDB) and clinical ECG data (NIFECGDB, PCDB) are used to evaluate the extraction performance of the proposed framework. The statistical and visual results demonstrate that our method outperforms the other state-of-the-art algorithms in the literature. Specifically, on the FECGSYNDB, the mean squared error (MSE), signal-to-noise ratio (SNR), correlation coefficient (R) and F1-score of our method are 0.16, 7.94, 0.95 and 98.89%, respectively. The F1-score on the NIFECGDB reaches 98.62%. The value of the F1-score on the PCDB is 98.62%. Conclusion As opposed to the existing algorithms being restricted to fetal QRS complex detection, the proposed framework can preserve the morphological features of the extracted fetal ECG signal well, which could support medical diagnoses based on the morphology of the fetal ECG signal.

Objective: Microvascular invasion (MVI) is an independent factor for postoperative recurrence of hepatocellular carcinoma (HCC). Accurate preoperative prediction of MVI grading is helpful for surgical planning in HCC management. We aimed to investigate the consistency and diagnostic performance of Magnetic resonance imaging(MRI) in assessing the presence of MVI, and the validity of deep learning attention mechanisms and clinical features in MVI grade prediction. Method: A total of 93 patients were selected from the Shunde Hospital affiliated to Southern Medical University in China. Retrospective image data and clinical data (n=93, collected between January,2017 and February,2020) were used to establish single sequence deep learning models and fusion models based on the EfficientNet and attention modules. Among them, the image data is enhanced by conventional MRI sequences (T1WI, T2WI, DWI), enhanced MRI sequences (AP, PP, EP, HBP) and synthesized MRI sequences (T1mapping-pre, T1mapping-20min). Furthermore, high-risk areas of hepatocellular carcinoma microvascular invasion were visualized by deep learning visualization techniques. Result: The fusion model based on T1mapping-20min sequence and clinical features outperforms other fusion models. Accuracy:0.8376; Sensitivity:0.8378; Specificity: 0.8702; AUC:0.8501. And deep fusion models can display MVI high-risk areas. Conclusion: Fusion model based on multiple MRI sequences and were successfully established. The effectiveness of deep learning algorithm was verified combined with attention mechanism and clinical features for MVI grading prediction. Therefore, the combination of deep attention mechanism and clinical features is an effective tool for preoperative prediction of MVI, which has advantages over using only deep features and radiomics.