Functional lumen imaging probe (FLIP) panometry was developed for real time assessment of esophagogastric junction opening and esophageal body contractile response during index endoscopy. Despite its proven value in functional esophageal disorders and its potential to reshape diagnostic pathways in Neurogastroenterology, substantial heterogeneity persists in catheter selection, procedural protocols and exam interpretation. In 2025 the Dallas Consensus provided standardized guidance on indications, performance, interpretation and reporting, yet exam interpretation remains complex and largely confined to high expertise centers. Artificial intelligence offers a pragmatic opportunity to simplify interpretation and support broader adoption of FLIP panometry in lower volume settings. This study aimed to develop an AI model capable of automatic interpretation of motility patterns during FLIP panometry according to the Dallas Consensus, with the goal of enabling more consistent and accessible use in clinical practice.

123 exams from 5 centers from both the European and American continents and performed with both 8- and 16-centimeter catheters were used for model development. Each procedure was classified with a two-expert based consensus decision according to the latest Dallas Consensus. Several machine learning models were trained and tested for evaluation of the esophagogastric junction opening and esophageal body contractile response. Data was divided in training and testing datasets with a 80%/20% patient split design, using 5-fold cross validation in the training dataset. Models’ performance was evaluated with the testing dataset trough their accuracy and area under the receiver-operating characteristic curve (AUC-ROC).

Disorders of esophagogastric junction opening were detected by CatBoost with 96.6% accuracy and an AUC-ROC of 0.989. Regarding disorders of contractile response, they were detected by XGBoost with 78.2% accuracy and an AUC-ROC of 0.913. Pathological planimetry patterns according to Dallas Consensus were identified by LightGBM with a mean accuracy of 92.1% and AUC-ROC of 0.955.

This study represents the first application of AI analysis to FLIP panometry, incorporating data from multiple probe types and heterogeneous demographic contexts while adhering to the most recent Dallas Consensus criteria, demonstrating robust accuracy in identifying pathological planimetry patterns. Subsequent research will refine these findings through dedicated validation efforts and explore the capacity to discriminate among distinct motility abnormalities. The integration of AI into FLIP panometry has the potential to meaningfully advance the diagnostic evaluation of suspected esophageal disorders during endoscopy by improving interpretative precision, reinforcing procedural standardization and expanding access beyond specialized centers, thereby contributing to more timely and effective patient management.