AI-Aided Colonoscopy May Improve Colorectal Neoplasia Detection

A systematic review and meta-analysis of randomized clinical trials (RCTs) published last week in eClinicalMedicine showed that artificial intelligence (AI)-aided colonoscopy may enhance the detection of colorectal neoplasia.

The work sought to assess the potential advantages and disadvantages of leveraging AI-aided systems during colonoscopies. The researchers noted that the use of these technologies to help detect colorectal neoplasia has the potential to improve adenoma detection rates (ADRs) and reduce adenoma miss rates (AMRs). However, results have been mixed across studies.

To compare AI-assisted with standard colonoscopy for detecting colorectal neoplasia, the research team used Medical Subject Headings (MeSH) terms and keywords to conduct an electronic literature search within the Embase, Medline, and the Cochrane Library databases. These databases were queried to identify relevant randomized controlled trials from the inception of each database until October 4, 2023.

The primary outcomes assessed were AMR, ADR, and adenomas detected per colonoscopy (APC), while secondary outcomes included the polyp missed detection rate (PMR), polyp detection rate (PDR), and polyps detected per colonoscopy (PPC).

The analysis included 33 RCTs with 27,404 total patients.

The research team found that patients undergoing AI-aided colonoscopy experienced a significant decrease in PMR and AMR — 52.5 percent and 50.5 percent, respectively — compared to their counterparts who received a standard colonoscopy.

Further, the researchers observed a significant increase in PDR, ADR, PPC, and APC rates in the AI-aided colonoscopy group. ADR and PDR saw a relative increase of 24.2 and 23.8 percent, while APC and PPC saw an increase of 39 and 38.8 percent.

These increases resulted in 0.271 more PPCs and 0.202 more APCs flagged on average. Mean inspection time also increased by 20 seconds.

The study underscored that certain populations may benefit more from AI-aided colonoscopy, such as younger patients with lower body mass index (BMI), as well as endoscopists with lower ADR or PDR and shorter inspection times.

Factors like time of day, anesthesia, and bowel preparation had significant influence on the efficacy of AI-aided colonoscopy in the RCTs reviewed.

These findings highlight that AI-aided colonoscopy may significantly improve the detection of patients with advanced adenomas, particularly those with non-neoplastic lesions. The use of AI tools may also drive significant increases in the detection of small adenomas, which could lead to enhanced surveillance strategies and reduce the risk of interval colorectal cancer (CRC).

Despite these promising results, the research team emphasized that more studies are needed to evaluate the cost-effectiveness and long-term benefits of leveraging AI-aided colonoscopy to reduce cancer incidence.

Other research has also investigated the potential of AI to improve CRC care.

Mayo Clinic-led research shared last year demonstrated that a deep learning (DL) framework could enhance the prediction of recurrence and survival in CRC patients.

Accurately predicting cancer recurrence in CRC is crucial for improved patient outcomes and survival rates, but making these predictions relies on multiple factors, presenting a challenge for clinicians. To combat this, the researchers developed a model to predict cancer recurrence using tumor images.

The model achieved high predictive performance, accurately identifying patients who may or may not need intensive treatment.