| Author | Message | ||
     Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 84 Registered: 2-2007 |
Jim: It�s been years since I looked at Tsuchikawa�s work, and it might be all changed by now, but I don�t think it should be termed a �Scatter Correction�, and I notice that that phrase does not show up in your last post. Your original post said that the PhD student was advised to consider 'improving' his modeling by using a Tsuchikawa scatter correction model - an extension to Kubelka-Munk theory. I think that the phrase invites people to consider applying his methods to data for which it is not suited, and in my opinion, that would be the case for data collected using �conventional NIRS�. As I remember his work described in Korea, he introduces various parameters and functions into the analysis of the TOF data. Some of the parameters were a multiplicative factor to account for absorptivity, a divisor to account for level of scatter, and what seemed to me to be a relatively arbitrarily inserted parameter which was different for various wavelength ranges. Then he related, as a function of wavelength, the reciprocal of that complex factor to the time delay for the maximum intensity. I thought it was beautiful work, but I didn�t conclude that he was modeling what he said he was (and I tried unsuccessfully to argue the point). But you Chemometric guys don�t care about that. You just want the method �validated�. I think in the end, the technique is an innovative way of introducing non-linear factors into an analysis in a way that avoids over fitting. The form of the function is being constrained by various factors that have an origin in theory. To some extent, the same thing could be said about MSC, but it is far more applicable to conventional spectroscopic data. It would be nice to get the folks who actually developed these schemes to participate in a discussion in some forum. If not here, suppose that might be possible in France? Don | ||
| Tony Davies (td) Moderator Username: td Post Number: 292 Registered: 1-2001 |
Hello Jim, Please could you send the reference for that Tsuchikawa paper? Thanks, Tony | ||
| Jim Burger (jburger) Member Username: jburger Post Number: 15 Registered: 11-2010 |
Great comments everyone! Thank you! I think I'm asking the (possibly) impossible (but never say impossible)- but it is a nagging issue, this scattering. Unfortunately we don't have access to any TOF like hardware, only conventional NIRS and a low (spectral) resolution hyperspectral instrument that at least does provide additional spatial resolution components. One of Tsuchikawa's papers summarizes his TOF-NIR Kubelka-Munk model extensions as: 1) Directional characteristics model 2) Light-path model 3) Equivalent surface roughness model 4) Generalized input/output equations for radiation My basic question is how can these models be extended when the light source is a diffuse source, not a highly parallel and uniform incident angle one. Jim | ||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 83 Registered: 2-2007 |
Don�t be troubled! You understand the situation. If we have a lightly scattering liquid, there could indeed be a substantial amount of light coming straight through. (Look at Gerry Birth�s photos of milk scatter on page 11 of Phil and Karl�s Food and Ag �Handbook�.) Samples like you mentioned will have almost no intensity at the minimum time and will rise up rather sharply beyond that, forming a peak that then tails off more slowly. It is the exact shape of this peak that is analyzed (using one's favorite model) to obtain the coefficients. There is some argument about whether the observed times are reasonable, and some folks content they show that the �photons� spend a lot of time at an absorption site waiting to be scattered, but I stay out of that fight. If we had measurable data at the minimum time, it�s extinction would follow the Bouguer-Lambert law and the observed extinction coefficient would be the sum of the absorption and scatter coefficient. | ||
| David W. Hopkins (dhopkins) Senior Member Username: dhopkins Post Number: 223 Registered: 10-2002 |
Hi Don, I find a couple of statements in your analysis of of light scattering and absorption in a transmission experiment troubling. I suspect I would need to have a better understanding of the sample model being considered, and the data analysis model, to evaluate the situation. You say, "The light which is scattered takes a longer path through the sample than the light which is not scattered." And further, "If we consider a transmission experiment, there is only one path through the sample with the minimum time. That is the path straight through the sample." It seems to me that in any real situation, such as the measurement of a grain sample in a cuvette or a section of wood, there will be no ray of light that is not scattered. Yet, there should be an observation of a minimum time. The question is, what use is made of that data. It would appear that it would be a mistake to assume that the minimum TOF involves no scattering, and therefore allows the estimation of absorption parameters alone. I am not familiar with the work of Tsuchikawa, so I would like to follow up on his results. Best regards, Dave | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 513 Registered: 9-2001 |
Tony - From the list of papers at the link to pubmed in my prvious message, many of Tsuchikawa's papers are in Appl. Spect. In fact comparing that to JNIRS, the score is: Appl. Spect = 11 JNIRS = 0 So if you can only access JNIRS, you are very limited as to which of Tsuchikawa's work you can read. However, that list also includes a few papers in miscellaneous other journals, if you go to the full list, you may find some that you have access to. I also note that a couple of the papers are coauthored with Heinz Siesler, so you could ask Heinz to send you copies of those papers. \o/ /_\ | ||
| Tony Davies (td) Moderator Username: td Post Number: 291 Registered: 1-2001 |
Hello Don, Been waiting for you to come in! This seems to me to be rather like "Norris Derivative" which was not given that name by Karl Norris. Has Satoru ever called it "Tsuchikawa scatter correction"? Even the web cannot find names that have not been used! As an editor of the proceedings of NIR-2001 (Korean ICNIRS conference) I was aware of Satoru's work on wood with ToF but Jim did not mention ToF so I did not make that connection. I looked at several of his recent JNIRS papers but did not see any useful references. Jim, will Don's answer satisfy your student's advisor? Best wishes and Happy Christmas to all our users Tony | ||
| Donald J Dahm (djdahm) Senior Member Username: djdahm Post Number: 82 Registered: 2-2007 |
Tsuchikawa has several TOF papers (one on wood) that are in the Proceedings of the Korean Conference. I tried to argue with him then, but my poor hearing and his accent made it impossible to have a productive discussion. There was half a session devoted to TOF at the �Physics� session at IDRC-2010 (which I chaired). Dave Burns had indeed presented some TOF stuff earlier in the conference. He has shown a video of light diffusing down along and below a tissue surface a few times there, as I recall. In truth, we have never gotten enough of the right people in the room at the same time to make much progress on this issue at a Conference. In the typical experiment, a pulse of light is sent and the intensity recorded at one or more positions as a function of time. The light which is scattered takes a longer path through the sample than the light which is not scattered. In theory, all light that reaches a detector at a particular time has taken a path of the same length. However, some of the light has taken a path with more scattering events with shorter distances between, and some with fewer scattering events with longer path segments between them. If we consider a transmission experiment, there is only one path through the sample with the minimum time. That is the path straight through the sample. The more times the light is scattered, the longer the path. Furthermore, on longer paths (and time), there are more pathways that have the same length. If we assume that a sample can be described by a single absorption and a single scattering coefficient, the pattern of intensity which comes through the sample should be resolvable into two patterns: a scattering one and an absorption one. The normal assumption is that the scattering curve must be a smooth monotone decreasing function of wavelength, while the absorption curve shows its characteristic ups and downs. (I am working on a paper in which it will be demonstrated that the scatter curve can have a maximum.) What I have been trying to point out (and have done so with varying degrees of distain in NIR News) that the scattering curve you obtain by most of these methods will depend on the absorption. What we are seeking, as a function of wavelength, is the scattering curve in the absence of absorption. A common technique in spectroscopy is �base line correction�, and if one can find a wavelength where there is no absorption nearby an absorption peak, the scatter correction should be straightforward. However, the complex systems (such as tissue) to which the technique is often applied may have no such regions. In my opinion, we are a ways from a practical theoretical solution, but there are several schemes that are good enough to be trying. But be prepared to tinker, and guide your tinkering with sound theory. | ||
| David W. Hopkins (dhopkins) Senior Member Username: dhopkins Post Number: 222 Registered: 10-2002 |
Hi Howard, I think I help you on one of your questions -- I believe it was Dave Burns who presented some TOF measurements, and it was 2010, I think. Or he might have mentioned earlier, too. Dave, can you set us right? Best regards, Dave | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 512 Registered: 9-2001 |
Well, Jim, I realized al long time ago that one of the most futile exercises you can do is to try to figure out why other people do what they do. Now you've made me realize there's something even more futile: to figure out why they DON'T do something!! But it's not hard to make a guess, in this case: The hardware is complicated, delicate and finicky (and don't forget expensive). And the software even more so. Most people in the NIR world, who are using it simply as a tool for conducting other studies, are intimidated and confused by as "simple" an algorithm as MLR. They just want a black box that gives answers. So putting all that together, you can count the number of people willing to brave all those factors, on the hands of one finger. I seem to recall there was a talk at Chambersburg, two or three meetings ago, where TOF NIR was presented. It was in a medical context; measurement of blood flow in the brain through the skull. If anyone could remember when it was, or who presented it, I might be able to come up with something about it. So there might be research for some medical applications, but published in places we don't normally check. There might be research for some military applications, which are SECRET so we can't access that info. I's not hard to come up with possibilities, but it's anyone's guess as to which are true. \o/ /_\ | ||
| Jim Burger (jburger) Member Username: jburger Post Number: 14 Registered: 11-2010 |
Hi Tony and Howard, Yes, the internet knows all. Or something approaching all (at least on 12.12.12). I'm very much aware of Satoru Tsuchikawa's publications on this theory, beginning with work at least as old as 1996. But that's my concern: I can't find ANY references to this work, except for what he has done. Is anyone else picking up on it? If not, why not? Jim | ||
| Howard Mark (hlmark) Senior Member Username: hlmark Post Number: 511 Registered: 9-2001 |
Tony - we seem to keep forgetting: the Internet knows EVERTHING!! Here are some results froa a search: Rapid prediction of past climate condition from lake sediments by near-infrared (NIR) spectroscopy: http://pubget.com/paper/22732538/Rapid_prediction_of_past_climate_condition_from_lake_sediments_by_near_infrared__NIR__spectroscopy There are a whole slew of links at: http://www.ncbi.nlm.nih.gov/pubmed?cmd=search&db=pubmed&term=Tsuchikawa+S%5Bau%5D&dispmax=50 Here's an excerpt from those listings, the one at "6." looks like it might be particularly pertinent: Near-infrared spectroscopic investigation of the hydrothermal degradation mechanism of wood as an analogue of archaeological wood. Part II: hardwood. Inagaki T, Mitsui K, Tsuchikawa S. Appl Spectrosc. 2009 Jul;63(7):753-8. doi: 10.1366/000370209788701071. PMID: 19589212 [PubMed - indexed for MEDLINE] Related citations 6.Application of time-of-flight near-infrared spectroscopy to fruits: analysis of absorption and scattering conditions of near-infrared radiation using cross-correlation of the time-resolved profile. Kurata Y, Tsuchikawa S. Appl Spectrosc. 2009 Mar;63(3):306-12. PMID: 19281646 [PubMed - indexed for MEDLINE] Related citations 7.Near-infrared spectroscopic investigation of the hydrothermal degradation mechanism of wood as an analogue of archaeological objects. Part I: softwood. Inagaki T, Mitsui K, Tsuchikawa S. Appl Spectrosc. 2008 Nov;62(11):1209-15. doi: 10.1366/000370208786401563. PMID: 19007461 [PubMed - indexed for MEDLINE] Related citations 8.Near-infrared spectroscopic monitoring of the water adsorption/desorption process in modern and archaeological wood. Inagaki T, Yonenobu H, Tsuchikawa S. Appl Spectrosc. 2008 Aug;62(8):860-5. doi: 10.1366/000370208785284312. PMID: 18702858 [PubMed - indexed for MEDLINE] Related citations | ||
| Tony Davies (td) Moderator Username: td Post Number: 290 Registered: 1-2001 |
Hello Jim You were just ahead of the "12 event". Your message was timed at 12:02,12/12/12. But I'm not sure the basis (may be Easter USA time?) Anyway about your question: This is new to me so it is either very old or very new (Tomas Hirschfeld's description of useful references). I haven't found it either. I believe that Tsuchikawa is a fairly common name in Japan but the likely one is Satoru Tsuchikawa. Perhaps you would like to ask him. He has many papers in JNIRS. Best wishes, Tony | ||
| Jim Burger (jburger) Member Username: jburger Post Number: 13 Registered: 11-2010 |
Hello All, I am assisting a new PhD student who is looking at moisture content measurement with NIR hyperspectral imaging. He was advised to consider 'improving' his modeling by using a Tsuchikawa scatter correction model - an extension to Kebulka-Monk theory. I was surprised that a search on this forum produced zero discussion of the Tsuchikawa model. Does anyone have any experience / comments to share regarding this model, or any other corrections to consider for wood samples? Thanks, Jim |