Too bad you rushed it in the end... I'm familiar with Michelson interferometery but you really went to quickly on the reason why I clicked on this video: explaining exactly how we build the image layer by layer...
how do we adjust the mirror for the next layer?in this video for example the green layer.I understood for L1=L2 we see blue but how do we see for green layer? Thnaks in advance
When we send red and orange light together, will the beamsplitter split the red light to object (or mirror) and the orange light to mirror (or object) ?
Great video! I really understand TD-OCT now, but could you explain how FD-OCT works? i can't really grasp how that would be implemented in this system. Do you detect all photons with different frequencies? How would different path lengths affect these numbers? I hope you read this!
Hi Zach. Decent very beginner video but I think it may be misleading to most because you are talking about versions of OCT which are no longer used. All current systems work in the spectral domain, require long coherence length, and have no scanning mirror. Most modern ones sweep the laser source to achieve an artificially large bandwidth and fine resolution. The A-scan information is recovered from the fourier transform of the fringes. To me the most difficult thing to grasp is the relation between bandwidth, coherence autocorrelation, and resolution... i.e. why sweeping through multiple wavelengths gives you higher resolution.
I am just learning about this, but I think the speaker misspoke at around 18:30. He said the coherence has to do with the difference between L1 and L2. Actually, the coherence length cares about the magnitudes of L1 and L2. As long as both L1 and L2 are less than the coherence length, you can still make reasonable observations regarding interference when the are varied by some small delta, provided that L2+delta is less than the coherence length as well.
+Charlie Freundlich Hey Charlie, thanks for the comment. I agree that I wasn't very clear on this point, so hopefully I can help with the confusion. The coherence length actually doesn't specifically deal with the L1 or L2 distances. Coherence in the context of light really deals with the ability of light to interfere. For instance, using a Michelson interferometer in the case of a laser (a perfectly coherent source) we can send half the beam a very large distance (ie L1=10m) and the second part a very short distance (ie L2=3mm) and there will still be interference because the laser uses light of very narrow bandwidth (the delta lamda in the equation at 20:15). However, when we have a light source with a broader bandwidth (with orange light as well as red) the coherence length shortens. In that situation there is only interference when the path length difference between the two legs is smaller than the coherence length. All this said, if L1 and L2 are shorter than the coherence length you are correct, there will be interference. Generally speaking, L1 and L2 are much longer than the coherence length which might only be a couple of microns. Thanks again! Let me know if there is further confusion.
also the fringe contrast depends upon the magnitude of intensity values reflected from the two arms, one can adjust it by using OD filters or by putting a glass slide instead of a mirror assuming that you sample is on a glass slide too in the other arm if you don't want to use ODs to complicate the system
@@zachnadler1914 I was also really confused about this and had the same thought that L1 and L2 had to be shorter than the coherence length. But I get it now. Even though L1 and L2 are larger than the coherence length and the individual light beams may not be coherent anymore, the phase shifts between the red and orange waves are still (almost) the same for L1 and L2 as long as the path length difference is within the coherence length. As long as the phase shift between red and orange is the same for the interfering beams, an interference pattern will emerge.
Great video. I have a couple of follow up questions. 1) I'm not sure I understood why the addition of "extra" frequencies (larger bandwidth light source) solved the picosecond laser/monochromatic issue (i.e. that picosecond lasers are too expensive). Is it just that a pulsed light source that isn't as spatially coherent as a monochromatic laser is much cheaper and that you can get away with a light source that has a reasonable coherent length? 2) Let's say you want to get L1 to equal L2. I imagine you can just calibrate until you see a bright spot. L1 might not be equal to L2 but you know that they are integer half wavelength in difference. Is this correct? Or are you able to move these within <400-700nm precision? Now let's say you're imaging a tissue and you don't know the tissue thickness a priori. When you want to get a second "slice" or cross-section, do you just move the mirror a pre-defined distance and see what you get or do you scan continuously and see what you get? Also what is moving the mirror such a fine distance? Stepper-motors or maybe some threaded rods? Thank you again.
Viraj Shah Great questions! Not sure I can adequately answer them in a comment but, I will try... 1. I was a little confused on this question. The broader the bandwidth of the light source equals a smaller the coherence length (and by extension finer axial resolution). Interference of light between the two legs of the interferometer only occurs when the path length of light traveled by one leg is within one coherence length of the distance traveled by the other leg. In regards to the femtosecond laser, one of the primary limitations I didn't mention is that the sensitivity of these sources are not great enough for biological tissues and can be 1000x less than the broad bandwidth sources. 2. Because the interference occurs within a coherence length you can bring the optical path lengths of the two legs to within tens of microns of one another just by observing interference and technically even closer than that. My engineering background is limited so I don't have a great answer for you in terms of the mechanism for scanning. Probably somewhat system dependent but galvos, resonant scanners and stepper motors sound good to me. One thing I left out is that the moving reference mirror was used in the original systems (called Time Domain OCT) but newer technology (Spectral Domain OCT) leverages a fourier transform to replace the moving reference mirror. These systems are faster and less susceptible to noise so most common commercial systems use SDOCT. Maybe I'll make a video detailing that technology. Thanks for the questions! Let me know if I can help any further!
Unfortunately the fourier part and building layer for layer is missing.
This litterally the best video i saw on youtube to understand what OCT is, in a more physical way. Thank you soooo much!!
i love you
Thank you!
does the L1 in the explanation correlate to depth and L2 width?
Great video - thanks !
Amazing! thank you for sharing with us.
V helpful
Thank you,it's a great video
Nice video! You explain the principales of OCT much better than my teachers hhh
amazing!! god bless you
Thank you.
very educative. thank you regards
Thanks a lot. It is a gerat introduction for OCT.
Simplified !! Good work. Thanks too.
Thank you very muck this really did help me a lot❤
can u do a video explaining the OCT FD and OCT time domain?
Can I have the references for this presentation....which article have you taken ???
Very good
Too bad you rushed it in the end... I'm familiar with Michelson interferometery but you really went to quickly on the reason why I clicked on this video: explaining exactly how we build the image layer by layer...
What a perfect mouse control!
how do we adjust the mirror for the next layer?in this video for example the green layer.I understood for L1=L2 we see blue but how do we see for green layer? Thnaks in advance
Nicely explained! Thanks Zach!
Painfully slow presentation. I recommend watching at 1.75x normal speed.
Thanks for the tip!
Brilliant! Subscribed.
Great tutorial Zach. Thank you!
Come on!!! Its not mikelson interferometer its michael + son its michelson intrf...
i love you
Thank you regards
Very informative
If we mention the 4th case for L1 and L2 distances, L1 is not equal L2, but the difference between them is not lamda/2. What happens?
When we send red and orange light together, will the beamsplitter split the red light to object (or mirror) and the orange light to mirror (or object) ?
outstanding and to the point. well done zach
great lecture
Should high myopes have one every year
Great explanation. Could you talk about the Spectral Domain OCT?
Great Video! Could you make one on Swept Source OCT?
Great video! I really understand TD-OCT now, but could you explain how FD-OCT works? i can't really grasp how that would be implemented in this system. Do you detect all photons with different frequencies? How would different path lengths affect these numbers? I hope you read this!
Thanks for sharing.
I have a doubt. Is it better to use light with high coherence length?
If it is high coherence length you don't get as much "spread" of the light and you won't see as much, but it would increase resolution
Thank you for the tutorial, it has helped me a lot since it was kind of hard to understand this written on a book.
Saved me a lot of trouble. Thanks a lot.
awesome explanation thank you!!!
Hi Zach. Decent very beginner video but I think it may be misleading to most because you are talking about versions of OCT which are no longer used. All current systems work in the spectral domain, require long coherence length, and have no scanning mirror. Most modern ones sweep the laser source to achieve an artificially large bandwidth and fine resolution. The A-scan information is recovered from the fourier transform of the fringes. To me the most difficult thing to grasp is the relation between bandwidth, coherence autocorrelation, and resolution... i.e. why sweeping through multiple wavelengths gives you higher resolution.
Max, do you have a link that explains how the current systems work?
I am just learning about this, but I think the speaker misspoke at around 18:30. He said the coherence has to do with the difference between L1 and L2. Actually, the coherence length cares about the magnitudes of L1 and L2. As long as both L1 and L2 are less than the coherence length, you can still make reasonable observations regarding interference when the are varied by some small delta, provided that L2+delta is less than the coherence length as well.
+Charlie Freundlich Hey Charlie, thanks for the comment. I agree that I wasn't very clear on this point, so hopefully I can help with the confusion. The coherence length actually doesn't specifically deal with the L1 or L2 distances. Coherence in the context of light really deals with the ability of light to interfere. For instance, using a Michelson interferometer in the case of a laser (a perfectly coherent source) we can send half the beam a very large distance (ie L1=10m) and the second part a very short distance (ie L2=3mm) and there will still be interference because the laser uses light of very narrow bandwidth (the delta lamda in the equation at 20:15). However, when we have a light source with a broader bandwidth (with orange light as well as red) the coherence length shortens. In that situation there is only interference when the path length difference between the two legs is smaller than the coherence length. All this said, if L1 and L2 are shorter than the coherence length you are correct, there will be interference. Generally speaking, L1 and L2 are much longer than the coherence length which might only be a couple of microns. Thanks again! Let me know if there is further confusion.
also the fringe contrast depends upon the magnitude of intensity values reflected from the two arms, one can adjust it by using OD filters or by putting a glass slide instead of a mirror assuming that you sample is on a glass slide too in the other arm if you don't want to use ODs to complicate the system
@@zachnadler1914 I was also really confused about this and had the same thought that L1 and L2 had to be shorter than the coherence length. But I get it now. Even though L1 and L2 are larger than the coherence length and the individual light beams may not be coherent anymore, the phase shifts between the red and orange waves are still (almost) the same for L1 and L2 as long as the path length difference is within the coherence length. As long as the phase shift between red and orange is the same for the interfering beams, an interference pattern will emerge.
Great tutorial :)
Great video. I have a couple of follow up questions. 1) I'm not sure I understood why the addition of "extra" frequencies (larger bandwidth light source) solved the picosecond laser/monochromatic issue (i.e. that picosecond lasers are too expensive). Is it just that a pulsed light source that isn't as spatially coherent as a monochromatic laser is much cheaper and that you can get away with a light source that has a reasonable coherent length? 2) Let's say you want to get L1 to equal L2. I imagine you can just calibrate until you see a bright spot. L1 might not be equal to L2 but you know that they are integer half wavelength in difference. Is this correct? Or are you able to move these within <400-700nm precision? Now let's say you're imaging a tissue and you don't know the tissue thickness a priori. When you want to get a second "slice" or cross-section, do you just move the mirror a pre-defined distance and see what you get or do you scan continuously and see what you get? Also what is moving the mirror such a fine distance? Stepper-motors or maybe some threaded rods? Thank you again.
Viraj Shah Great questions! Not sure I can adequately answer them in a comment but, I will try... 1. I was a little confused on this question. The broader the bandwidth of the light source equals a smaller the coherence length (and by extension finer axial resolution). Interference of light between the two legs of the interferometer only occurs when the path length of light traveled by one leg is within one coherence length of the distance traveled by the other leg. In regards to the femtosecond laser, one of the primary limitations I didn't mention is that the sensitivity of these sources are not great enough for biological tissues and can be 1000x less than the broad bandwidth sources. 2. Because the interference occurs within a coherence length you can bring the optical path lengths of the two legs to within tens of microns of one another just by observing interference and technically even closer than that. My engineering background is limited so I don't have a great answer for you in terms of the mechanism for scanning. Probably somewhat system dependent but galvos, resonant scanners and stepper motors sound good to me. One thing I left out is that the moving reference mirror was used in the original systems (called Time Domain OCT) but newer technology (Spectral Domain OCT) leverages a fourier transform to replace the moving reference mirror. These systems are faster and less susceptible to noise so most common commercial systems use SDOCT. Maybe I'll make a video detailing that technology. Thanks for the questions! Let me know if I can help any further!
Zach Nadler This is great and very helpful. Thank you!
Nice work Zach! We're looking forward to great things!
Great stuff!!!