algorithm for counting peaks

Dr. Ross Lazarus rossl at gmu.wh.su.edu.au
Wed Aug 3 21:15:08 EST 1994

There's a whole raftload of literature on peak identification and 
characterization in noisy data. I'm a little familiar with the problem in 
the context of endocrine research into pulsatile hormone activity - 
typical of hormones such as LH and growth hormone in most species.

Suffice it to say the problem is far from trivial ! The shifting baseline is 
a particular challenge. I've implemented two of the most commonly used 
algorithms (Veldhuis' CLUSTER and Merriam & Wachter's PULSAR) in a pc 
package which can cope with fairly horrible data - be warned, it's a 
black art. I'd be happy to discuss availability and pricing via direct email 
with anyone interested...

In article <31m741$sjq at oak.oakland.edu> mje at pookie.pass.wayne.edu (Michael J. Edelman) writes:
>From: mje at pookie.pass.wayne.edu (Michael J. Edelman)
>Subject: algorithm for counting peaks
>Date: 2 Aug 1994 19:34:25 GMT

>I've been asked to write a program that will read time-series data consisting
>of counts of events per period, and output the numbers of peaks in a given 
>specified time interval. The Tricky part is in defining what constitutes a peak. 

>The researchers are using some canned package from Data Translation that doesn't 
>perform this task; they've been eyeballing the data up until now. It's been 
>about 16 years since I did any work in electrophysiology where I used to do
>this kind of coding, so I'm at a loss where to start. 

>Can someone direct me to a good reference, or perhaps suggest a simple algorithm?
>We may also have to deal with a shifting baseline in the data as well.

Dr Ross Lazarus, Head, Epidemiology and Biostatistics Unit,
Community Medicine, Westmead Hospital, Westmead NSW 2145, Australia.
Email: rossl at gmu.wh.su.edu.au  Fax: +61 2 6891049  Tel: +61 2 6336677

More information about the Bio-soft mailing list

Send comments to us at biosci-help [At] net.bio.net