An intelligent diagnosis and treatment system for in-hospital cardiac arrest based on deep reinforcement learning in the Utstein style

Abstract

Background: Both in-hospital and out-of-hospital cardiac arrest have several causes and complexities. Therefore, it is difficult to analyze and create targeted treatment plans for cardiac arrest. Moreover, even basic patient informationis insufficient or missing in many cases. To address these challenges, we developed an intelligent diagnosis and treatment system for cardiopulmonary resuscitation and restoration of spontaneous circulation to reasonably complete the diagnosis and treatment process and improve the success rate of treatment during in-hospital cardiac arrest. Methods: Using 10 years of data from Utstein style in-hospital cardiac arrest registries in seven hospitals in China’s Hebei Province, we developed a two-stage intelligent diagnosis and treatment system based on the basic situation of different cases during cardiac arrest using deep reinforcement learning. This approach is aimed at maximizing the success rates of cardiopulmonary resuscitation and maintaining blood pressure at an optimal value during the restoration of spontaneous circulation. We employed Monte Carlo simulation to analyze the uncertainty. Results: When comparing the actual diagnosis and treatment with those based on other models during cardiopulmonary resuscitation and restoration of spontaneous circulation in different conditions (sufficient and insufficient basic patient information), the average reward scores of the treatment plan provided by the intelligent diagnosis and treatment system developed in this study increased to 9.9–23 from 2.39. Conclusions: The intelligent diagnosis and treatment system for in-hospital cardiac arrest based on deep reinforcement learning under the Utstein mode developed in this study can propose a reasonable diagnosis and treatment plan when in-hospital cardiac arrest occurs. Moreover, it can effectively induce cardiopulmonary resuscitation and restoration of spontaneous circulation processes in different patients even when basic patient informationis insufficient or missing.

Machine learning in TCM with natural products and molecules: current status and future perspectives

“…Natural product development Deep neurol network [47,50], RF [53,54,58,59,93], SVM [51,53,54,57,59,93], DT [59,93], neural network [53,59] RF was better than SVM, neurol network and DT in screening hepatotoxic compounds [59]. RF model is more accurate than SVM and DT in identifying molecular characteristics of natural product compounds with the meridians of TCM [93] Disease diagnosis SVM [10,61,66,[81][82][83], DT [68,[81][82][83], neural network [45, 61-63, 65, 82, 83], RF [61,64,67,82,83], CNN [64,67,[70][71][72][73][74][75][76][77][78]81], RNN…”

“…With advancements in CNN, ResNet solved the problem of difficult training, high error rates, and a rapid decline in accuracy after the CNN depth increases. Shao [75] first separated the tongue and tongue coating, and then used the separated images as input to classify the tongue and tongue coating using ResNet-50. Residual connections make the CNN deeper, stronger, and more efficient.…”