The study revealed that the integration of a real-time computer-aided polyp detection system during colonoscopy substantially improves efficacy by increasing adenoma and polyp detection rates, as well as enhancing accuracy in terms of sensitivity and specificity. Recent research has further established that artificial intelligence-assisted colonoscopy (AIC) serves as a safe and efficient screening tool, improving the detection rates of colorectal cancer adenomas and polyps in adults and facilitating early diagnosis of colorectal cancer when compared to conventional colonoscopy [18, 19].

Colorectal cancer screening methods are effective but have limitations, and many eligible individuals are unscreened. This paper reviews current and new screening techniques and suggests ways to overcome these limitations. While new strategies could revolutionize prevention, significant improvements can be made by optimizing current colonoscopy methods. Colonoscopy diagnostics have a rich history [33].

Over the past decades, alternative methods such as CT colonography, colon capsule endoscopy, and AI-assisted detection have been explored to address the limitations of traditional gastrointestinal endoscopy [34, 35]. While auxiliary inspection techniques can enhance detection rates, the sensitivity of white light colonoscopy often fails to identify invasive flat or depressed lesions, which may progress to larger, advanced tumors. Fluorescence molecular imaging addresses this shortcoming by significantly improving the detection rate of precancerous lesions [36]. In a previous study [35], it was demonstrated that AI-assisted colonoscopy enhanced overall ADR, advanced ADR, and ADR among both expert and non-expert endoscopists. In addition, the implementation of a novel AI detection system resulted in a significantly higher number of adenomas detected per colonoscopy compared to conventional high-definition colonoscopy, without extending the colonoscopy withdrawal time. This finding supports the use of AI-assisted colonoscopy to enhance the quality of colonoscopy in a large, prospective, multicenter, randomized clinical trial [37].

As emerging technologies continue to develop, computer-aided science has been increasingly applied across various fields, particularly in medicine, to enhance patient quality of life. Gastrointestinal endoscopy, with its long-standing history in detecting, diagnosing, and treating colorectal lesions [33, 38], remains the most suitable test for high-risk individuals or as a follow-up procedure after a positive initial test [39]. Despite the current advancements in gastroenterology, there are perspectives that express reservations about the technological future of endoscopy. To optimize gastrointestinal endoscopy for reducing the risk of colorectal cancer, it is essential to refine our strategies to enhance the detection of adenomas and polyps [40].

Notably, our findings indicate an inverse relationship between the adenoma detection rate (ADR) and the risk of colorectal cancer at any stage. Furthermore, increasing polyp detection rates can significantly lower the incidence of colorectal cancer [21, 41, 42]. Consequently, we employed ADR and polyp detection rate (PDR) as key indicators in numerous studies to assess the effectiveness of a real-time computer-aided polyp detection system during colonoscopy [43]. Previous research has established a strong correlation between polyps and the onset and progression of colorectal cancer [23]. Therefore, to improve early-stage colorectal cancer detection, our objective is to elevate polyp detection rates, given that colorectal cancer is a genetic disease originating from precursor colon lesions or polyps that evolve through various tumorigenesis pathways [44]. Artificial intelligence (AI) has the potential to enhance polyp detection; however, numerous uncertainties persist. Despite minor variations among different AI algorithms, the overall efficacy of AI in polyp detection is promising, as evidenced by clinical trials [45]. Various strategies have been identified to improve the polyp detection rate [46, 47]. Nevertheless, a more critical issue lies in the limitations of algorithmic procedures. While CADe has been shown to increase adenoma detection rates in randomized trials, it still faces significant challenges, particularly in identifying certain pathological types of lesions. For instance, the first commercially available CADe program in the United States, GI Genius, demonstrates suboptimal performance in detecting large, flat lesions. Similarly, although the new Medtronic CADe program, ColonPRO, incorporates advanced AI capabilities, preliminary evaluations indicate that it continues to struggle with effectively detecting certain lesions, especially after excluding specific patient groups. Certain lesions necessitate modifications in colonoscopy procedures to elicit signals, whereas others may not generate any signals. The efficacy of CADe is contingent upon the training of the algorithm. If the algorithm is primarily trained to identify small lesions, it may fail to detect large, flat lesions, potentially increasing the ADR but not effectively preventing colorectal cancer. Thus, it is imperative to balance the algorithm to detect all types of lesions for effective colorectal cancer prevention [48]. A meta-analysis indicates that AI-assisted colonoscopy can substantially enhance the adenoma detection rate; however, regardless of the endoscopists’ experience, system type, or medical environment, it does not improve the detection rate of sessile serrated lesions. This suggests inherent limitations within the artificial intelligence algorithm, aligning with our statistical findings, where external factors were not statistically significant [49].

In this prospective trial, we analyze and compare the adenoma detection rate (ADR) and polyp detection rate (PDR) between standard colonoscopy and computer-aided colonoscopy to evaluate the efficacy and monitor the safety of the system during the procedure. The study aims to assess the accuracy and safety of the computer-assisted diagnostic software for detecting intestinal polyps via gastrointestinal endoscopy, building upon previous research involving colonoscopy screening with fecal occult blood tests [50]. Our study employs the polyp per colonoscopy (PPC) as the primary outcome measure, a metric routinely recorded during gastrointestinal polyp endoscopy, thereby enhancing the feasibility of data collection. Preliminary data indicate that software-assisted gastrointestinal polyp endoscopy can significantly increase PPC. This trial is informed by prior exploratory clinical research, providing valuable experience in project design, training, implementation, and statistical analysis, which enhances the probability of success.

Researchers have employed various approaches to investigate the potential of artificial intelligence (AI) in enhancing the detection of adenomas and polyps, which holds significant implications for the prevention of colorectal cancer [51]. The methodology of our study parallels that of a randomized trial, which prospectively assessed the use of a computer-aided detection device across five academic and community centers by gastroenterologists certified by the US board [52]. In both studies, participants were randomly assigned to two groups to evaluate whether AI assistance could improve the detection rates of adenomas and polyps. However, while the previous study assessed safety based on the severity of adverse events and their relationship to the intervention, our study evaluated safety by examining the rates of bleeding and perforation. In addition, we drew insights from a prior trial that explored AI-assisted adenoma detection [53].

While the results align with the hypothesis, the study has several flaws, notably the lack of a double-blind design. This trial examines PPC differences between experimental and control groups, primarily influenced by the number of polyps and endoscopists’ detection abilities. Extremely low or high polyp numbers or a very low missed diagnosis rate could lead to negative outcomes. However, the study’s population, individuals consecutively needing endoscopy, should have a moderate polyp count. In addition, previous research indicates even experienced physicians have some missed diagnoses, reducing the likelihood of trial failure.

These findings suggest a promising avenue for the prevention of colorectal cancer by emphasizing the advancement of artificial intelligence technologies. AI holds substantial potential in the creation of innovative screening strategies designed to enhance the detection of adenomas and polyps, thereby contributing to an improved quality of life. The application of AI across diverse medical domains has yielded favorable outcomes, significantly enhancing diagnostic and detection rates, and consequently improving overall quality of life [54].

Future research should consider employing double-blind designs to more accurately evaluate the specific role of the system in enhancing adenoma detection rates, as the observed behavior may influence these rates in the experimental group. Given that baseline adenoma and polyp detection rates may be affected by regional characteristics, the study’s findings may not be applicable to regions with higher baseline adenoma detection rates globally. Further investigation is required to assess the system’s adaptability and effectiveness in such regions. Research should focus on directions such as improving processing speed while maintaining high accuracy to meet the demands of real-time examinations. In addition, large-scale, multicenter randomized controlled trials are necessary to validate the effectiveness of artificial intelligence systems in clinical settings. Customizing models based on the specific characteristics of different hospitals and equipment could enhance treatment outcomes.

In conclusion, this study demonstrated the efficacy and accuracy of the software in real-time colonoscopy and identified the sensitivity and specificity of the system.