Functional analysis of oesophageal clearance from 24 hour ambulatory pH recordings comparing patients with Barrett’s oesophagus and healthy controls

This work takes a new look at pH signals and extracts functional information regarding swallowing performance. First clearance curves are extracted from 24 hour pH recordings these are then fitted to a formal model of clearance. These are then compared between between two groups. A web publication (Copyright Haylett, 2011) can be downloaded here:

Functional analysis of oesophageal clearance from 24 hour ambulatory pH recordings comparing patients with Barrett’s oesophagus and healthy controls, Haylett 2011.

Example: This shows a web based example of the basic technique developed within this work.

Analysis of ambulatory oesophageal manometry

My latest research has also involved investigating and improving techniques to analyse ambulatory oesophageal manometry. Ambulatory manometry was first introduced in the mid eighties but has gradually gone out of fashion. Initially there was great enthusiasm for this technique. Standard laboratory based oesophageal manometry had been and still is problematic. With just a small window of observations how do the results relate to day to day swallowing activity and symtoms? The results from laboratory studies are still problematic and the relationship between observations and diseases is by far from clear. Terms and concepts are frequently introduced and there is much confusion about what value the results can be given. How should we interpret the results?

Typical computer analysis is not trusted for 24 hour ambulatory recordings (see page 5 for a simple model of reflux and page 6 for a review http://www.medeng.net/volume_9_4.pdf  ). This is largely because the waves are complex and the algorithms used are linear only detecting previously well defined patterns. Small signals and ‘noise’ can disrupt the analysis. Importantly anything not predefined is not captured by the basic peak detection algorithms.

The data mining approach

In our approach a simple data mining technique is used. This described in the following papers presentation.

http://www.medeng.net/IEE%20MASP.pdf and http://www.medeng.net/York%202007.pdf

In summary all possible events of interest are collected and then clustered using a Kohonen self organising feature map. These clusters can then be explored by the user. In addition post cluster classification can be carried out and results compared between subjects or groups of subjects

Examples

The following examples are copyright © K.R. Haylett 2008 and can only published or used with the consent of the author.

Example 1: This shows an example of analysing a 24 hour recording and the clusters created. It also shows post cluster classification into primary, secondary and non-peristaltic waveform classes.

Example 2: This shows a comparison between a series of controls and patients with Barrett’s oesophagus using post cluster classification.

The technique is MATLAB based and can be carried out online via the internet. For further information please contact me.