|
Author
|
Topic: blood pressure cuff vs. photoplethysmograph
|
duras
Member
|
posted 01-31-2007 07:07 PM
  
For a valid PDD examination to exist, respiration, ED, and cardio activity must be monitored and recorded. As you know, some instruments also record finger pulse amplitude using a photoplethysmograph. My question: Is it possible to use photoplethysmograph as main cardio sensor rather than blood pressure cuff (instead of, not in addition)? If it is not possible, please explain why. I really need this clarification for my Russian colleagues. Thank you very much.IP: Logged |
dkrapohl
Member
|
posted 01-31-2007 09:20 PM
The simple answer is no. The two channels are not interchangeable. Though the photoplethysmograph (PLE) and the cardio cuff both capture the pulse, the PLE detects the vasomotor response (capillary dilation) while the cuff detects systemic pressure changes. Ideally, examiners will record and score both since they provide diagnostic information that is independent of each other and the other polygraph channels. Mark Handler has submitted an article to Polygraph that will appear later this year. It tells all about the PLE, and why we should consider adding it to our polygraphs.IP: Logged |
duras
Member
|
posted 02-01-2007 03:00 AM
Sure, I totally agree with you on this subject. However, during the debate with my colleagues from Russia I have faced intensive criticism for this position. Did you know that many Russian polygraph examiners (including experienced professionals working for law enforcement agencies) use photoplethysmograph in preference to blood pressure cuff? This is a fact. And that's a huge amount of people. I failed to convince my Russian colleagues of the justice of my views. That is why I have posted this question here, in hopes of a response. I will look forward for the article you have mentioned. Thank you for your answer.IP: Logged |
Barry C
Member
|
posted 02-01-2007 09:44 AM
It's anecdotal I know, but I have used the PLE for a while now, and I can tell you the cardio channel captures a lot more than the PLE. In other words, you're going to miss reactions if you only use the PLE. (The reason can be explained physiologically, and I suspect we'll see it in Mark's article.) Less data translates to reduced accuracy, and why would you want to do that? Is it ethical? I don't think so.Have you shown them charts of the cardio vs PLE? I would think demonstrating that they can't possibly be recording the same data (and as Don explained, they don't) would be enough to persuade any rational person. This raises another issue I haven't yet posted, but will if people are interested. There is only one study I know of on finger (BP) cuffs vs arm cuffs, and that study (a DoDPI study) showed the two don't record the same data, which means we have no data to show the finger cuffs add anything positive to our tests either. IP: Logged |
dkrapohl
Member
|
posted 02-01-2007 12:20 PM
Anyone needing a copy of this report can find it reprinted in Polygraph (1997), 26,(2), 69-78. The report suggests that the researchers used the old thumb transducers versus the more recent finger cuff. Too early to say whether that would make any difference in the conclusions. It is certain that this cuff is different from the PLE. Barry's comment about having less diagnostic information in the PLE versus the cuff has been verified in research by Kircher and Raskin (1988) and possibly others. The examiners who use the PLE in place of a measure of systemic blood pressure have chosen a channel that delivers less diagnostic information. Such a practice would be unethical for members of professional organizations in the West.IP: Logged |
Barry C
Member
|
posted 02-01-2007 06:08 PM
I've got an electronic (PDF) version of the finger cuff study I can email to those who don't have the APA journal and would like to take a look at it. It's short, and the cuff is described enough to know it is different from what's being marketed today. Duras, Get a copy of the article Don referenced, and might I suggest you get the most recent copy of POLYGRAPH. Mark Handler discusses scoring of the cardio and PLE in explaining the Utah test, and he cites previous work (Bell et al, if my memory is correct - I have that one in PDF somewhere too) showing scores of other than 0 are much more common in the cardio channel than in the PLE channel. If they were both recording the same thing, then scoring should be consistent. That might make more sense to some, but who knows? It might also convince them that they could have fewer inconclusive tests. This strikes me as a strange debate as you're really talking apples and oranges. It seems odd that people wouldn't know that's the case. IP: Logged |
duras
Member
|
posted 02-01-2007 06:47 PM
Barry, could you please email me these PDF files? My address is as follows: sergiy.duras at gmail.com And please recommend me the most authoritative piece of writing on membrane plethysmograph and photoplethysmograph (technical description, scoring rules, etc). Thanks for your attention! IP: Logged |
Barry C
Member
|
posted 02-01-2007 09:06 PM
I sent you Bell et al, Kircher & Raskin, and the finger cuff study. I don't have Mark's article in PDF, but maybe he'll send you a copy (if you don't have the latest POLYGRAPH journal). The same info is found in Bell et al, but I think Mark made it a little more clear. The Kircher article is highly technical, so you've got something for everyone. Good luck.IP: Logged |
duras
Member
|
posted 02-02-2007 02:13 AM
Thank you very much indeed.IP: Logged |
skar
Member
|
posted 01-23-2011 05:55 AM
Effects of Prior Demonstrations of Polygraph Accuracy on Outcomes of Probable-Lie and Directedlie Polygraph Tests. John C. Kircher and others, 2001.
"Table 4. Validity (and reliability) of differential reactivity indices for PL tests
Cardiograph - .45 (.56)
Finger Pulse Amplitude - .43 (.54)
Table 5. Validity (and reliability) of differential reactivity indices for DL tests
Cardiograph - .36 (.60)
Finger Pulse Amplitude - .53 (.59)"Maybe I misunderstood something but why we can`t replace arm cuff with PLE if it has such validity and reliability? Are there big differences among validities and reliabilities or something else? Thanks. p.s. I am using both arm cuff and PLE. [This message has been edited by skar (edited 01-23-2011).] IP: Logged |
Barry C
Member
|
posted 01-23-2011 02:09 PM
Well, if you read that study closely (or the 1988 computer algorithm study), you'll see that the PLE data didn't add anything to the model. That is the adjusted R-square was maximized with cardio - not PLE data. Thus, they aren't doing the same thing, which we already knew. Just because you have a high point-biserial correlation doesn't mean the variable will add to the model. Some info is redundant. (We did find a way to add PLE to the traditional model in which the adjusted R-square was (significantly) maximized, but that's a different issue.)IP: Logged |
skar
Member
|
posted 04-16-2011 07:35 AM
But why the PLE data with such validity and reliability didn't add anything to the model? I don`t know what is adjusted R-square. Was it because Kircher and others simply decided to do that? Or there were objective reasons? What were the reasons in this case?
Thanks.[This message has been edited by skar (edited 04-16-2011).] IP: Logged |
Barry C
Member
|
posted 04-16-2011 08:13 PM
Well, I just put in a nice explanation, but it disappeared somehow.In any event, it was objective. The way Kircher measured the PLE, even though it highly correlated with deception, it didn't increase accuracy of the algorithm. You probably could have used the PLE in place of RLL and been pretty close, but the RLL+EDA+CARDIO model worked the best - and no better with PLE. Ben Blalock and I (and Kircher and Bernhardt) have found a way to measure it in such a way as to have it add to the R-square (i.e., add useful information to the model / algorithm), and we're working on the paper now. It was Don Krapohl's idea to measure it like we do RLL, and we now measure "FPLL" - and it works. It's just taken me a long time to do the analysis. I had to learn how to bootstrap, run simulations, etc, and I wanted to be sure it was right first. If I can, I'll post some of the accuracy info here. I'm trying to get my part of writing finished so we can get it published. I had hoped to be done this weekend, but it may be another weekend before I am happy with it. The story of my life.... IP: Logged |
Barry C
Member
|
posted 04-16-2011 08:33 PM
Okay, let's see if I can remember how to do this:
The CIs are computed using the exact method, except the LRs, for which I used the Agresti-Coull method. (If I didn't, we ended up with CIs that extended beyond 1 at times. The stats geeks will understand that. If you don't, don't worry about it.) IP: Logged |
rnelson
Member
|
posted 04-18-2011 07:41 AM
Barry and all,I think this might actually be a finger-pulse excursion measurement, and not really line length. The difference is important. Either way, the next massive challenge is to build it into a normed decision model... and this will require at least two samples of representative data to get us to where we are at the present time without the pulse-oximiter. Increasing the criterion validity of the polygraph is not simply a matter of adding another component. The polygraph, at the present time, is so accurate (~.90) that it would be possible to accomplish no improvement, and possibly even decrease decision effectiveness, if we were to go about this in an unscientific manner. Generalizable improvements in norm-referenced criterion accuracy beyond what we can presently achieve with OSS-3 and ESS will be expensive (in terms of study and research hours), and will require thorough and careful statistic analysis. Anyway, we're working on it. r ------------------
"Gentlemen, you can't fight in here. This is the war room."
--(Stanley Kubrick/Peter Sellers - Dr. Strangelove, 1964)
IP: Logged |
Barry C
Member
|
posted 04-18-2011 07:55 AM
It is (excursion). That's the way CPS-LAB measures it, and it is important as it solves the time problem.IP: Logged |
rnelson
Member
|
posted 04-18-2011 06:18 PM
barry,are the CIs in your table confidence intervals for the mean score or the normal range of scores? Also, is there any way you could recalculate that table with the inconclusive and error rates separated for the truthful and deceptive cases? r ------------------
"Gentlemen, you can't fight in here. This is the war room."
--(Stanley Kubrick/Peter Sellers - Dr. Strangelove, 1964)
IP: Logged |
Barry C
Member
|
posted 04-19-2011 09:16 AM
The CIs here are all CIs for the proportions calculated, e.g., overall accuracy, errors, etc. - not the scores. I have the breakdowns of TPs, FPs, etc, for everything, including the INCs, but I don't have the data here (with me right now). I'll try to remember to post it tonight. (This table may not be clear, but it wasn't intended for the paper. It's just a portion of summary info I made for myself to make sure I get a good picture of what the data reveals.)The beauty of the MC is the precision of the estimates. While you don't see a significant difference between the with and without FP in the training and validation samples (88 and 56 cases, respectively), you do see the small increase with the simulated data, which is what we see with the regression model with the original data, even with only 88 cases. While I won't be reporting on it (I don't think), I looked at the raw data for the validation sample and had the computer simulate scores using a modified version of ESS in which the FP got a -1,0, or +1 and the others, the regular ESS scores. If the measurement differences were greater than 1.1 to 1, I had Excel assign the appropriate score. The results were good. I don't recall them off the top of my head, but they were around 90%. IP: Logged |
Barry C
Member
|
posted 04-19-2011 09:32 PM
Here's the INC breakdowns for the MC validation sets:With FPLL Validation Set:
T-INCs: 384 (7.68%) [95% CI: 6.96%|8.45%]
D-INCs: 322 (6.44%) [95% CI: 5.78%|7.16%] No FPLL Validation Set:
T-INCs: 481 (9.62%) [95% CI: 8.82%|10.47%]
D-INCs: 694 (13.88%) [95% CI: 12.93|14.87%] [This message has been edited by Barry C (edited 04-20-2011).] IP: Logged | |
Polygraph Place Bulletin Board
