Machine Learning Tool Helps Guide Treatment for Esophagogastric Junction Obstruction

A diagnostic tool with machine learning can help provide diagnostic decision support for esophageal obstruction.

A new diagnostic tool based on probabilistic machine learning can aid in diagnosis and clinical decision-making following functional lumen imaging probe (FLIP) studies for patients with esophagogastric junction obstruction (EGJO), according to findings published in the Journal of Neurogastroenterology and Motility.

Researchers at Northwestern University developed a diagnostic tool using machine learning that aimed to apply FLIP data to accurately diagnose EGJO compared with high-resolution manometry (HRM).

Researchers undertook this study based on previously performed prospective cohort studies between November 2012 and December 2019 to assess how well their diagnostic tool worked.

The previous studies collected data from 557 adult patients between the ages of 18 and 89 years old who underwent assessments using FLIP during upper endoscopies as well as HRM. The investigators compared data from these 557 symptomatic patients with data from 35 healthy, asymptomatic individuals (controls).

Probabilistic machine learning offers a promising solution to alleviate the limitations associated with classification schemes using fixed rules and thresholds for clinical diagnoses.

Following HRM, 243 patients obtained confirmation of EGJO according to the Chicago Classification version 4.0 (CCv4.0) diagnostic criteria. The remaining 314 patients and all 35 controls demonstrated normal esophagogastric junction outflow.

Following two different analyses (leave-one-out and holdout test set), the diagnostic model predicted the presence of EGJO with 89% and 90% accuracy, respectively.

According to the two analyses, the diagnostic tool demonstrated 87% and 85% sensitivity and 92% and 97% specificity for diagnosing EGJO, respectively.

 “Probabilistic machine learning offers a promising solution to alleviate the limitations associated with classification schemes using fixed rules and thresholds for clinical diagnoses,” the study authors wrote. “Probabilistic machine learning offers a promising solution to alleviate the limitations associated with classification schemes using fixed rules and thresholds for clinical diagnoses. Future directions will seek to incorporate additional metrics or features (such as the contractile response patterns on FLIP Panometry) into the model to improve statistical precision and clinical performance.”

Study limitations include those inherent to supervised training labels of the proposed diagnostic model and the lack of a perfect standard of measurement of EGJO.

Disclosures: Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.

References:

Schauer JM, Kou W, Prescott JE, Kahrilas PJ, Pandolfino JE, Carlson DA. Estimating probability for esophageal obstruction: A diagnostic decision support tool applying machine learning to functional lumen imaging probe panometry. J Neurogastroenterol Motil. 2022;28(4):572-579. doi:10.5056/jnm21239