A multi-stage machine learning model for diagnosis of esophageal manometry

作者：

Highlights：

• A multi-stage artificial intelligence model for esophageal manometry was developed.

• Three convolutional neural network models on swallow type, pressurization and IRP were built.

• Study-level models to predict esophageal motility diagnoses were then developed.

• A model-agnostic approach of model balancing was proposed to develop blended models.

• The proposed framework is extensible for multi-modal tasks with multiple data sources.

摘要

•A multi-stage artificial intelligence model for esophageal manometry was developed.•Three convolutional neural network models on swallow type, pressurization and IRP were built.•Study-level models to predict esophageal motility diagnoses were then developed.•A model-agnostic approach of model balancing was proposed to develop blended models.•The proposed framework is extensible for multi-modal tasks with multiple data sources.

论文关键词：High-resolution manometry,Artificial intelligence,Model averaging

论文评审过程：Received 23 June 2021, Revised 17 December 2021, Accepted 18 December 2021, Available online 25 December 2021, Version of Record 3 January 2022.

论文官网地址：https://doi.org/10.1016/j.artmed.2021.102233