Gastric Cancer
Conditions
Brief summary
Our AIMING project comprises four core work packages (WPs): WP1. Nation-level randomized controlled trial; WP2. Development of an innovative AI tool; WP3. Novel microsimulation modelling; WP4. Patient inclusion. The nation-level multi-center tandem randomized controlled trial (WP1) will contribute to a better understanding of how the real-time AI algorithm can reduce miss rate of early gastric cancer and dysplasia during gastroscopy. Moreover, the innovation project will contribute to development of a novel AI tool (WP2) that can stratify the risk of gastric cancer by identifying in vivo precancerous conditions. Furthermore, a microsimulation modelling will allow us to predict how the use of AI can prevent gastric cancer and affect cost and patients' burdens. The assessment of the balance between benefits and harms is quite crucial especially for this type of medical device because the value of innovative tools is sometimes overestimated due to stakeholders' enthusiasm (WP3). Finally, we will take care of patients' perspective throughout the study project by including patient organization in both WP1, 2, and 3 (WP4).
Interventions
Two novel deep learning systems, namely one for endoscopy and one for pathology, will be trained and validated for the diagnosis of gastric atrophy and metaplasia, including extension and severity. Both of the algorithms will be validated against the cases not used for the training phases. Approximately, the partition will be 5 to 1. The benefit and harm of AI-assistance for early diagnosis of gastric cancer will be simulated by developing a Markov model on the natural history of gastric cancer from dysplasia to early and advanced cancer, as well as by the impact of a GS on its natural history. This will also simulate the potential effect of lead- and length-time bias. These data will be incorporated in the simulation model in order to include them in the decision-making process on whether AI-assistance for gastric cancer detection should be or not recommended to health systems.
Sponsors
Istituto Clinico Humanitas
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
PREVENTION
Masking
NONE
Intervention model description
Parallel/Crossover Study Model; Patients will be randomized 1:1:1:1
Eligibility
Sex/Gender
ALL
Age
60 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* All \>60 years-old patients undergoing upper-gastrointestinal (GI) endoscopy for selected indications at high-risk of gastric cancer..
Exclusion criteria
* contraindications to upper-GI endoscopy. * contraindications to biopsy. * active upper-GI bleeding or urgent upper-GI endoscopy. * patients with previous upper-GI surgery involving the stomach. * patients who were not able or refused to give informed written consent.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Miss rate reduction | 2025: 12 months enrollment | change of the miss rate of early gastric cancer and dysplastic lesions at upper-endoscopy when using AI-assistance (tandem). |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Change number of Detections | 1 day procedure and follow up for 2 years | Change in the detection of early gastric cancer and dysplastic lesions at upper-endoscopy when using AI-assistance (parallel). |
| patient satisfaction | 2025: during the 12 months enrollment | Assessment of patient acceptability, satisfaction and tolerance, assessed by questionnaire, towards AI technology for both the detection and the characterization of gastric lesions. |
Outcome results
None listed