Aim. This study proposes a new artificial intelligence model based on cardiovascular computed tomography for more efficient and precise recognition of Tetralogy of Fallot (TOF). Methods. Our model is a structurally optimized stochastic pooling convolutional neural network (SOSPCNN), which combines stochastic pooling, structural optimization, and convolutional neural network. In addition, multiple‐way data augmentation is used to overcome overfitting. Grad‐CAM is employed to provide explainability to the proposed SOSPCNN model. Meanwhile, both desktop and web apps are developed based on this SOSPCNN model. Results. The results on ten runs of 10‐fold crossvalidation show that our SOSPCNN model yields a sensitivity of 92.25 ± 2.19, a specificity of 92.75 ± 2.49, a precision of 92.79 ± 2.29, an accuracy of 92.50 ± 1.18, an F1 score of 92.48 ± 1.17, an MCC of 85.06 ± 2.38, an FMI of 92.50 ± 1.17, and an AUC of 0.9587. Conclusion. The SOSPCNN method performed better than three state‐of‐the‐art TOF recognition approaches.

Working memory (WM) – one of the brain ability that maintains information – can evaluate the function of brain. Activities related to memory sustention, inhibition and disinhibition have gathered significant attention for the basic neurocognitive architecture. Although researchers have proposed some brain models that attempt to explain the entire procedure of WM, little evidence can proof and describe it, and more particularly, regions and structures of maintenance, inhibition and disinhibition require more investigation. We used phase lock coherence and general partial directed coherence to construct connections among four adaptively fitted EEG sources, and we also applied previous published models to describe the brain circuits of maintenance, inhibition and disinhibition. Referring to a classical visual n-back paradigm, we recruited forty five mental health undergraduates in this experiment. We found that the bilateral prefrontal cortex (PFC) mainly focused on some cognitive components, for example, rehearsal before recognition to classify objects, inhibition to maintain positive memory and activities, and disinhibition to arouse or activate subsequent interactions in brain. Meanwhile, the right PFC sometimes could assist left PFC to implement high capacity WM tasks. By contrast, the posterior regions, PPC, tends to be engaged in attention arousing and maintaining. These two findings suggest that a) the recurrent maintenance circuit may keep the brain executing positive cognitive components, b) then the instantly monitoring inhibition would pause the deadlocked sustention function to save energy, and c) the arriving of disinhibition arouses the next step in brain to select new subject or focus on novel subjects.

Adversarial attacks dramatically change the output of an otherwise accurate learning system using a seemingly inconsequential modification to a piece of input data. Paradoxically, empirical evidence indicates that even systems which are robust to large random perturbations of the input data remain susceptible to small, easily constructed, adversarial perturbations of their inputs. Here, we show that this may be seen as a fundamental feature of classifiers working with high dimensional input data. We introduce a simple generic and generalisable framework for which key behaviours observed in practical systems arise with high probability-notably the simultaneous susceptibility of the (otherwise accurate) model to easily constructed adversarial attacks, and robustness to random perturbations of the input data. We confirm that the same phenomena are directly observed in practical neural networks trained on standard image classification problems, where even large additive random noise fails to trigger the adversarial instability of the network. A surprising takeaway is that even small margins separating a classifier's decision surface from training and testing data can hide adversarial susceptibility from being detected using randomly sampled perturbations. Counter-intuitively, using additive noise during training or testing is therefore inefficient for eradicating or detecting adversarial examples, and more demanding adversarial training is required.