Ensemble deep learning for the prediction of proficiency at a virtual simulator for robot-assisted surgery

Background Artificial intelligence (AI) has the potential to enhance patient safety in surgery, and all its aspects, including education and training, will derive considerable benefit from AI. In the present study, deep-learning models were used to predict the rates of proficiency acquisition in robot-assisted surgery (RAS), thereby providing surgical programs directors information on the levels of the innate ability of trainees to facilitate the implementation of flexible personalized training. Methods 176 medical students, without prior experience with surgical simulators, were trained to reach proficiency in five tasks on a virtual simulator for RAS. Ensemble deep neural networks (DNN) models were developed and compared with other ensemble AI algorithms, i.e., random forests and gradient boosted regression trees (GBRT). Results DNN models achieved a higher accuracy than random forests and GBRT in predicting time to proficiency, 0.84 vs. 0.70 and 0.77, respectively (Peg board 2), 0.83 vs. 0.79 and 0.78 (Ring walk 2), 0.81 vs 0.81 and 0.80 (Match board 1), 0.79 vs. 0.75 and 0.71 (Ring and rail 2), and 0.87 vs. 0.86 and 0.84 (Thread the rings 2). Ensemble DNN models outperformed random forests and GBRT in predicting number of attempts to proficiency, with an accuracy of 0.87 vs. 0.86 and 0.83, respectively (Peg board 2), 0.89 vs. 0.88 and 0.89 (Ring walk 2), 0.91 vs. 0.89 and 0.89 (Match board 1), 0.89 vs. 0.87 and 0.83 (Ring and rail 2), and 0.96 vs. 0.94 and 0.94 (Thread the rings 2). Conclusions Ensemble DNN models can identify at an early stage the acquisition rates of surgical technical proficiency of trainees and identify those struggling to reach the required expected proficiency level.

PREDICTION OF FORMATION ENERGY USING TWO-STAGE MACHINE LEARNING BASED ON CLUSTERING

The formation energy (Hf) is one of the important properties associated with the thermodynamic stability of ABO3-type perovskite. In this work, two-stage machine learning based on hierarchical clustering and regression was designed for improving the prediction values of the density-functional theory (DFT) Hf of ABO3-type perovskites. A global dataset was clustered into Cluster 1 and Cluster 2 using the CHI (the Calinski-Harabasz index). To compare the prediction performances of Hf, DTR (decision tree regression), GBRT (gradient boosted regression trees), RFR (random forest regression) and ETR (extra tree regression) were applied to build models of Cluster 1, Cluster 2 and the global dataset, respectively. The results showed that all four different regression models of Cluster 1 had a higher R2, and lower MSE and MAE than those of the global dataset, while the models of Cluster 2 were poorer. Meanwhile, the GBRT model of Cluster 1 achieved a higher R2 of 0.917, and lower MSE and MAE of 0.033 eV/atom and 0.125 eV/atom. We further validated and compared the generalization ability of the models by predicting the Hf of ABO3-type perovskite previously unseen in the training set. The two-stage machine-learning models proposed here can provide useful guidance for accelerating the exploration of materials with desired properties.

Classification of fragile states based on machine learning

The study of fragile states has become a significant issue in global security, development and poverty at present. The existing classification methods of fragile state, which is a simple addition to the national index and threshold segmentation, is not reasonable enough. We introduce a new method based on machine learning. With this method, it will be easier and more reasonable to classify a country. We use two kinds of classifier, one of which is the support vector machine, and the other is the gradient boosted regression trees. Both models have flaws, so we use ensemble learning techniques to combine them. First of all, subjective labelling of a part of the national data to allows the machine to learn why a country becomes vulnerable from these data, and how to classify the vulnerability class of a country. Then, we trained the model with the data, and divided fragile states into four categories successfully (Alert, Warning, Stable and Sustainable). For the classification result, our model got a 93% test error rate, and a 96% training error rate, which is better than 77% with the threshold segmentation method.