JMIR Publications

ABSTRACT

Background:

Background:

The onset time of acute ischemic stroke, of great significance in clinical medicine, is hard to estimate with existing tools (e.g., MRI). Machine learning might be a potentially effective method to identify and predict stroke onset time, whereas its application value remains controversial and debatable.

Objective:

Object: To map and assess the application value of machine learning in judging the time of stroke onset.

Methods:

Methods:

PubMed, Web of Science, Embase, and Cochrane were comprehensively searched. The c-index, sensitivity with 95% confidence intervals were used as effect sizes. The risk of bias was evaluated using PROBAST and meta-analysis was conducted in R4.2.0.

Results:

Results:

The systematic search yielded 13 eligible articles, 55 machine learning models with 41 models in the training set and 14 in the validation set. The overall c-index was 0.800 (95% CI: 0.773~0.826) in the training set and 0.781 (95% CI: 0.709~0.852) in the validation set. The sensitivity and specificity in the train set and validation set were: sensitivity [train: 0.76 (95% CI: 0.73,0.80), validation: 0.81 (95% CI: 0.68, 0.90)], and specificity [train: 0.79 (95% CI: 0.74, 0.82), validation: 0.83 (95% CI: 0.73, 0.89)].

Conclusions:

Conclusion: Results indicate the ideal performance of machine learning to identify the time of stroke onset. However, the prediction accuracy still needs improve, future research should focus on reasonable image segmentation and texture extraction methods in radiomics. Furthermore, promotion and application value of machine learning should be also investigated in different ethnic backgrounds.