OBJECTIVE: Degenerative phenomena may affect esophageal motility as a relevant social-health problem. The diagnosis of such disorders is usually performed by the analysis of data from high-resolution manometry (HRM). Inter- and intraobserver variability frequently affects the diagnosis, with potential interpretative and thus therapeutic errors, with unnecessary or worse treatments. This may be avoided with automatic procedures that minimize human intervention in data processing. METHODS: In order to support the traditional diagnostic process, an automatic procedure was defined considering a specific physiomechanical model that is able to objectively interpret data from HRM. A training set (N = 226) of healthy volunteers and pathological subjects was collected in order to define the model parameters distributions of the different groups of subjects, providing a preliminary database. A statistical algorithm was defined for an objective identification of the patient's healthy or pathological condition by comparing patient parameters with the database. RESULTS: A collection of HRMs including subjects of the training set has been built. Statistical relationships between parameters and pathologies have been established leading to a preliminary database. An automatic diagnosis procedure has been developed to compare model parameters of a specific patient with the database. The procedure was able to match the correct diagnosis up to 86% of the analyzed subjects. CONCLUSION: The success rate of the automatic procedure addresses the suitability of the developed algorithms to provide a valid support to the clinicians for the diagnostic activity. SIGNIFICANCE: The objectivity of developed tools increases the reliability of data interpretation and, consequently, patient acceptance.
OBJECTIVE: Degenerative phenomena may affect esophageal motility as a relevant social-health problem. The diagnosis of such disorders is usually performed by the analysis of data from high-resolution manometry (HRM). Inter- and intraobserver variability frequently affects the diagnosis, with potential interpretative and thus therapeutic errors, with unnecessary or worse treatments. This may be avoided with automatic procedures that minimize human intervention in data processing. METHODS: In order to support the traditional diagnostic process, an automatic procedure was defined considering a specific physiomechanical model that is able to objectively interpret data from HRM. A training set (N = 226) of healthy volunteers and pathological subjects was collected in order to define the model parameters distributions of the different groups of subjects, providing a preliminary database. A statistical algorithm was defined for an objective identification of the patient's healthy or pathological condition by comparing patient parameters with the database. RESULTS: A collection of HRMs including subjects of the training set has been built. Statistical relationships between parameters and pathologies have been established leading to a preliminary database. An automatic diagnosis procedure has been developed to compare model parameters of a specific patient with the database. The procedure was able to match the correct diagnosis up to 86% of the analyzed subjects. CONCLUSION: The success rate of the automatic procedure addresses the suitability of the developed algorithms to provide a valid support to the clinicians for the diagnostic activity. SIGNIFICANCE: The objectivity of developed tools increases the reliability of data interpretation and, consequently, patient acceptance.
Authors: Wenjun Kou; Dustin A Carlson; Alexandra J Baumann; Erica N Donnan; Jacob M Schauer; Mozziyar Etemadi; John E Pandolfino Journal: Artif Intell Med Date: 2021-12-25 Impact factor: 5.326