Introduction
Videocolonoscopy (VCC) remains the gold standard in the diagnosis of colorectal carcinoma (CRC) and premalignant lesions. Around 26% of polyps are missed during conventional colonoscopy due to operator (e.g., expertise, fatigue, etc.), and technical (e.g., suboptimal bowel preparation and withdrawal time) dependent-factors. Artificial intelligence (AI) aids in adenoma and polyp detection, thus increasing adenoma and polyp detection rates (ADR and PDR, respectively).
Aims & Methods
We aim to evaluate the real-world effectiveness of an AI-assisted colonoscopy (AI-VCC) and compare the results among experts and non-experts endoscopists. Consecutive patients ≥45 years old who underwent VCC (Nov/2020-Mar/2021) were invited to participate. Patients with BBPS≤6, history of inflammatory bowel disease, familial polyposis syndrome, colorectal carcinoma, colorectal surgery, or uncontrolled coagulopathy were not considered. The study protocol was conducted according to the declaration of Helsinki and approved by the local Institutional Review Board (IRB). All patients signed an informed written consent. First, a standard HD-VCC was performed. Then, an AI-VCC was performed by a second operator blinded to previous findings. We determined the adenoma detection rate (ADR) as a quality indicator of the procedure. The polyp detection rate (PDR), and the adenoma and polyp miss rates (AMR and PMR, respectively) were also calculated. The results were compared between experts and non-experts endoscopists. Data was analyzed in R v4.0.
Results
A total of 115 patients were included. Median age was 57.0 (51.0 - 67.5) years, 82/115 (71.3%) were females. 58/115 (50.4%) patients had a total of 205 polyps/adenoma, with a PDR and ADR of 132/205 (64.3%) and 73/205 (35.6%) with VCC, respectively: 19/205 (9.3%) NICE II, 7/205 (3.4%) >10 mm. With AI-assistance, the ADR and PDR increased from 16.5% to 18.2% and from 50.4% to 60%, respectively. According to the level of expertise, the ADR increased from 10.8% to 16.2% in the junior group with the AI-VCC, whereas no difference in ADR was observed in the senior cohort (Table 1). The calculated PMR was 43.13% while the AMR was 5.19%. Sensitivity among different colon segments ranged from 56.25% (rectum) to 78.26% (ascending colon). Meanwhile, capability of AI for appropriate polyp and adenoma presence discharging (negative predictive value, NPV) ranged from 61.54% (sigmoid) to 89.09% (cecum). However, lower rates for specificity (8.42-49.49%) and positive predictive value (PPV, 9.89-19.78%) were observed due to high false positives. (Table 2).
Table 1. ADR and PDR during HD-VCC and AI-VCC in senior and junior endoscopists.
| Standard VCC
| Secondary
AI-assisted VCC
| Standard +
AI-assisted VCC
| Missed Lesions Cases
|
Total (n=115; 100%)
|
ADR
PDR
| 19 (16.5)
58 (50.4)
| 6 (5.2)
33 (28.7)
| 21 (18.2)
69 (60.0)
| 2
11 |
Senior (n=78; 67.8%)
|
ADR
PDR
| 15/78 (19.2)
39/78 (50.0)
| 3 (3.8)
23 (29.5)
| 15/78 (19.2)
49/78 (62.0)
| 0
10 |
Junior (n=37; 32.2%)
|
ADR
PDR
| 4/37 (10.8)
19/37 (51.1)
| 3 (8.1)
10 (27.0)
| 6/37 (16.2)
20/37 (54.0)
| 2
1 |
ADR, Adenoma detection rate; PDR, polyp detection rate; HD-VCC, high-definition video colonoscopy; AI-VCC, Artificial intelligence-assisted video colonoscopy
|
Conclusion
The AI-assisted polyp detector is a feasible tool to aid endoscopists during screening colonoscopy. Additionally, with AI-assistance, junior endoscopists can achieve an ADR comparable to that of senior endoscopists. NCT04915833.
References
- Wang P, Berzin TM, Glissen Brown JR, Bharadwaj S, et al. Real-time automatic detection system increases colonoscopic polyp and adenoma detection rates: a prospective randomized controlled study. Gut. 2019 Oct;68(10):1813-1819.
- Zhao S, Wang S, Pan P, Xia T, Chang X, et al. Magnitude, Risk Factors, and Factors Associated with Adenoma Miss Rate of Tandem Colonoscopy: A Systematic Review and Meta-analysis. Gastroenterology. 2019;156 (6):1661–1674e11.
- Mahmud N, Cohen J, Tsourides K, Berzin TM. Computer vision and augmented reality in gastrointestinal endoscopy. Gastroenterol Rep (Oxf). 2015 Aug;3(3):179-84.
- Kröner PT, Engels MM, Glicksberg BS, Johnson KW, et al. Artificial intelligence in gastroenterology: A state-of-the-art review. World J Gastroenterol 2021; 27(40): 6794-6824.
- Taghiakbari M, Mori Y, von Renteln D. Artificial intelligence-assisted colonoscopy: A review of current state of practice and research. World J Gastroenterol. 2021 Dec 21;27(47):8103-8122.
Disclosure
Carlos Robles-Medranda is a key opinion leader and consultant for Pentax Medical, Boston Scientific, Steris, Medtronic, Motus, Micro-tech, G-Tech Medical Supply, CREO Medical, EndoSound, and Mdconsgroup. The other authors declare no conflicts of interest.
