The Effectiveness of Artificial Intelligence-based Colonoscopy: Propensity Score Matching Study _x000D_
<br>_x000D_
<br>Adenomas, accounting for over 70% of colon cancer origins, make their detection during colonoscopy a critical procedure in colon cancer prevention. To enhance the quality of colonoscopy, in which a detection rate of over 25% for adenomas is recommended. Artificial intelligence based on deep learning system has been developed and is under study in countries such as the United States, Japan, and China. This study aims to evaluate the effectiveness of the domestically developed deep learning system-based artificial intelligence, Endoscopy as AI-powered Device(ENAD). This study retrospectively analyzes and compares the outcomes of the two groups undergoing colonoscopy with and without the assistance of ENAD. Out of 654 colonoscopy cases reviewed, 197 were excluded. 104 cases were assisted by ENAD, while 353 were not. Propensity score matching was used to reduce selection bias, matching 104 subjects in each group, followed by a comparative analysis of the two groups. The study conducted the Mann-Whitney U-test to analyze continuous dependent variables and the Chi-square test to compare frequencies for categorical dependent variables. the Wilcoxon’s signed rank test to analyze continuous dependent variables and the McNemar’s test to compare frequencies for categorical dependent variables after the propensity score matching. All statistical analyses were conducted using SPSS Statistics version 29. Regarding the baseline characteristics, no significant differences were observed between the two groups in terms of age, female ratio, Body Mass Index (BMI; kg/m²), Obesity ratio, two categories of the American Society of Anesthesiologists (ASA) Score, Boston Bowel Preparation Scale, the indications of colonoscopy, and the proportion of inexperienced practitioners before and after the propensity score matching. In the post-colonoscopy outcomes, the ENAD assisted group showed higher average polyp detection rate and adenoma detection rate compared to the non-assisted group, but these differences were not statistically significant before and after the propensity score matching. The study concludes that the colonoscopy with the assistance of ENAD did not show a significant increase in adenoma detection rates and polyp detection rates compared to the colonoscopy without its assistance. Further research is necessary, involving additional analysis of some variables such as medical history, more detailed indications for colonoscopy, and a larger number of subjects. _x000D_
<br>Keywords: Deep Learning System, Artificial Intelligence, Colonoscopy, Adenoma Detection Rate, Polyp Detection Rate
Alternative Abstract
The Effectiveness of Artificial Intelligence-based Colonoscopy: Propensity Score Matching Study _x000D_
<br>_x000D_
<br>Adenomas, accounting for over 70% of colon cancer origins, make their detection during colonoscopy a critical procedure in colon cancer prevention. To enhance the quality of colonoscopy, in which a detection rate of over 25% for adenomas is recommended. Artificial intelligence based on deep learning system has been developed and is under study in countries such as the United States, Japan, and China. This study aims to evaluate the effectiveness of the domestically developed deep learning system-based artificial intelligence, Endoscopy as AI-powered Device(ENAD). This study retrospectively analyzes and compares the outcomes of the two groups undergoing colonoscopy with and without the assistance of ENAD. Out of 654 colonoscopy cases reviewed, 197 were excluded. 104 cases were assisted by ENAD, while 353 were not. Propensity score matching was used to reduce selection bias, matching 104 subjects in each group, followed by a comparative analysis of the two groups. The study conducted the Mann-Whitney U-test to analyze continuous dependent variables and the Chi-square test to compare frequencies for categorical dependent variables. the Wilcoxon’s signed rank test to analyze continuous dependent variables and the McNemar’s test to compare frequencies for categorical dependent variables after the propensity score matching. All statistical analyses were conducted using SPSS Statistics version 29. Regarding the baseline characteristics, no significant differences were observed between the two groups in terms of age, female ratio, Body Mass Index (BMI; kg/m²), Obesity ratio, two categories of the American Society of Anesthesiologists (ASA) Score, Boston Bowel Preparation Scale, the indications of colonoscopy, and the proportion of inexperienced practitioners before and after the propensity score matching. In the post-colonoscopy outcomes, the ENAD assisted group showed higher average polyp detection rate and adenoma detection rate compared to the non-assisted group, but these differences were not statistically significant before and after the propensity score matching. The study concludes that the colonoscopy with the assistance of ENAD did not show a significant increase in adenoma detection rates and polyp detection rates compared to the colonoscopy without its assistance. Further research is necessary, involving additional analysis of some variables such as medical history, more detailed indications for colonoscopy, and a larger number of subjects. _x000D_
<br>Keywords: Deep Learning System, Artificial Intelligence, Colonoscopy, Adenoma Detection Rate, Polyp Detection Rate
