Automatic Diagnosis of High-Resolution Esophageal Manometry Using Artificial Intelligence.

BACKGROUND AND AIMS: High-resolution esophageal manometry (HREM) is the gold standard procedure used for the diagnosis of esophageal motility disorders (EMD). Artificial intelligence (AI) might provide an efficient solution for the automatic diagnosis of EMD by improving the subjective interpretation of HREM images. The aim of our study was to develop an AI-based system, using neural networks, for the automatic diagnosis of HREM images, based on one wet swallow raw image. METHODS: In the first phase of the study, the manometry recordings of our patients were retrospectively analyzed by three experienced gastroenterologists, to verify and confirm the correct diagnosis. In the second phase of the study raw images were used to train an artificial neural network. We selected only those tracings with ten test swallows that were available for analysis, including a total of 1570 images. We had 10 diagnosis categories, as follows: normal, type I achalasia, type II achalasia, type III achalasia, esophago-gastric junction outflow obstruction, jackhammer oesophagus, absent contractility, distal esophageal spasm, ineffective esophageal motility, and fragmented peristalsis, based on Chicago classification v3.0 for EMDs. RESULTS: The raw images were cropped, binarized, and automatically divided in 3 parts: training, testing, validation. We used Inception V3 CNN model, pre-trained on ImageNet. We developed a custom classification layer, that allowed the CNN to classify each wet swallow image from the HREM system into one of the diagnosis categories mentioned above. Our algorithm was highly accurate, with an overall precision of more than 93%. CONCLUSION: Our neural network approach using HREM images resulted in a high accuracy automatic diagnosis of EMDs.

Automated Chicago Classification for Esophageal Motility Disorder Diagnosis Using Machine Learning.

The goal of this paper is to provide a Machine Learning-based solution that can be utilized to automate the Chicago Classification algorithm, the state-of-the-art scheme for esophageal motility disease identification. First, the photos were preprocessed by locating the area of interest-the precise instant of swallowing. After resizing and rescaling the photos, they were utilized as input for the Deep Learning models. The InceptionV3 Deep Learning model was used to identify the precise class of the IRP. We used the DenseNet201 CNN architecture to classify the images into 5 different classes of swallowing disorders. Finally, we combined the results of the two trained ML models to automate the Chicago Classification algorithm. With this solution we obtained a top-1 accuracy and f1-score of 86% with no human intervention, automating the whole flow, from image preprocessing until Chicago classification and diagnosis.

Integrated Relaxation Pressure Classification and Probe Positioning Failure Detection in High-Resolution Esophageal Manometry Using Machine Learning.

High-resolution esophageal manometry is used for the study of esophageal motility disorders, with the help of catheters with up to 36 sensors. Color pressure topography plots are generated and analyzed and using the Chicago algorithm a final diagnosis is established. One of the main parameters in this algorithm is integrated relaxation pressure (IRP). The procedure is time consuming. Our aim was to firstly develop a machine learning based solution to detect probe positioning failure and to create a classifier to automatically determine whether the IRP is in the normal range or higher than the cut-off, based solely on the raw images. The first step was the preprocessing of the images, by finding the region of interest-the exact moment of swallowing. Afterwards, the images were resized and rescaled, so they could be used as input for deep learning models. We used the InceptionV3 deep learning model to classify the images as correct or failure in catheter positioning and to determine the exact class of the IRP. The accuracy of the trained convolutional neural networks was above 90% for both problems. This work is just the first step in fully automating the Chicago Classification, reducing human intervention.

Edge computing for space applications: Field programmable gate array-based implementation of multiscale probability distribution functions.

Data processing is a challenging problem in space applications. The limited bandwidth available for communication between satellites and the ground and the increasing resolution of scientific instruments make it virtually impossible to transfer all the data recorded on board. Although various mitigation strategies were developed, large amounts of on-board data are still lost. This paper presents a Field Programmable Gate Array (FPGA)-based architecture which is able to perform on-board nonlinear analysis of data and compute probability distribution functions of fluctuations. We propose two implementations for our solution, which can be used for space applications and also other computational contexts. On a spacecraft, the logic resources of the FPGA will typically be shared by several designs running various digital signal processing algorithms. That is why each algorithm should be designed in variants, optimized for different criteria, so that the entire group of algorithms makes an efficient usage of the FPGA resources. The proposed solution focuses on two major optimization criteria, area and speed, such that the FPGA resources are efficiently used. Also, the power consumption is at least two orders of magnitude less in comparison with classical software implementations. The solution was tested with both synthetic and real data and shows excellent results paving the way towards an application that can be ported on a space-grade FPGA.