#458 - [Journal Club] - 📌 Could a handheld light probe detect NEC before the X-ray does?

NEC still tends to announce itself on an X-ray, usually later than we want. This week Daphna reviews the first-in-human study of broadband optical spectroscopy, a handheld transcutaneous probe out of Lurie Children's that reads reflected light from 350 to 2500 nanometers to look at bowel beneath the skin. Just under 100 preterm infants, more than 11,000 scans, and a machine learning model that separated NEC from non-NEC scans with 100% sensitivity and a 100% negative predictive value, though the positive predictive value sat at 24%. Ben and Daphna work through what the curves actually show and what would have to come next.

----

First-in-human pilot study of broadband optical spectroscopy (BOS) as noninvasive surveillance for necrotizing enterocolitis (NEC). Dodd AC, Lehane AJ, Lee A, Hurlock A, Su Y, Ilahi I, Lautz TB, Backman V, Goldstein SD.J Pediatr Surg. 2026 May;61(5):162978. doi: 10.1016/j.jpedsurg.2026.162978. Epub 2026 Feb 3.PMID: 41643767

Transcript

Ben Courchia, MD (00:00.59) Hello everybody, welcome back to the Incubator Podcast Journal Club. We're back. This is Thursday. Daphna has a very anticipated, highly anticipated paper to discuss with us.

Daphna Yasova Barbeau, MD (00:10.644) Highly anticipated. I like that. Yeah, the title is in it. It's called "First-in-Human Pilot Study of Broadband Optical Spectroscopy as Noninvasive Surveillance for Necrotizing Enterocolitis." This is actually in the Journal of Pediatric Surgery, so some people may not have seen it. Lead author Ashley Dodd, senior author Seth Goldstein.

Ben Courchia, MD (00:12.876) The title. The title is so puzzling.

Daphna Yasova Barbeau, MD (00:39.92) And so basically they wanted to look at, could we use broadband optical spectroscopy, or BOS, which we'll use as a shorthand, which is a transcutaneous non-invasive tool that has been studied, so people know, in mice, with a mouse model of necrotizing enterocolitis. Could we use it on babies to identify NEC?

And so let me give you a little background information. I think that will be helpful. So some people might be familiar with near-infrared spectroscopy, right? NIRS. So NIRS takes information like a venous oxygen saturation. It's really looking at the difference between oxygenated and deoxygenated blood, basically.

Ben Courchia, MD (01:18.071) Mm-hmm.

Daphna Yasova Barbeau, MD (01:32.556) And the tools that we use, the little sticker sensors, give us that change, that delta, as a readout on monitors, but it's specifically using that near infrared wavelength. And so this is somewhat different, because it's using a broad band of wavelengths, and it covers a very wide spectrum, so a spectrum between 350 and 2500 nanometers.

Ben Courchia, MD (01:53.826) Hmm.

Daphna Yasova Barbeau, MD (02:02.863) And it's a little different, not just in the spectrum of light that it's using, but how is it working? So basically you take the probe, and it provides a light source to your tissue of interest. And for example, they wanted to study the abdomen, and looking at

the intestines below the infant's skin. And then it reflects. So any light that you emit reflects some type of light pattern back. So this uses the light source and a spectrometer, which helps evaluate some of that light reflection, and so that's how it's a little bit different than NIRS testing. Does that make sense?

Ben Courchia, MD (02:58.742) It does.

Daphna Yasova Barbeau, MD (02:59.655) Okay, so basically you're able to use more wavelength spectra, so you can see the difference in reflectance at each wavelength. And so for different tissues and different disease states, there may be, and that's what they're looking for, a difference of reflectance across the wavelengths. You get a lot more information instead of just the information you might get from just one wavelength.

And as opposed to the oximeters we use for NIRS, they're looking at not just tissue oxygenation, which again, we know that our NIRS can be affected with tissue oxygenation, like say in the intestine, intestinal NIRS or cerebral NIRS. It can be from perfusion, it can be from inflammation, it could be from fluid like ascites.

So all of those things can impact tissue oxygenation, but it doesn't exactly tell us about the disease state that's causing the difference in oxygenation. So their hope is that with BOS, we can use these disease specific chromophores. So are there different spectral patterns in the disease pathology itself? So like in tissue

hypoxia versus tissue necrosis, will we get different spectral signals? So that's the background on BOS, and why they're looking at it as an alternative to, say, NIRS, which hasn't quite yet panned out as we had hoped to predict NEC. So I told you about how it works. I told you

how it works on the baby. They identified study participants, premature neonates, so 36 weeks gestation or fewer. And basically they followed them over the course of the NICU stay. They excluded babies with congenital heart disease and major abdominal wall defects like gastroschisis and omphalocele. And then they looked at the babies over the course of the study period.

Daphna Yasova Barbeau, MD (05:18.175) And also looked at which babies got NEC, so Bell stage two or three, and then they included 24 hours of documented pneumatosis or portal venous gas on X-ray. And then they looked at the differences in the BOS measurements. They looked at four abdominal quadrants, the epigastric, the right lower quadrant, the suprapubic, and the left lower quadrant,

at each encounter. And then they looked at the outer thigh, so the thigh muscle, as a control, since it's not viscera. This usually required one to two minutes during routine neonatal nursing care and was repeated two to three times weekly as long as the clinical status allowed. Or

as soon as possible, if there was a concern for a clinical status change, especially in the form of abdominal disease. And then basically, this is a machine learning paper, they were teaching the algorithm to detect differences in well babies versus these babies who had NEC. They had a total of 11,570 scans acquired.

196 of those scans, so still a pretty small proportion, are from infants who were diagnosed with NEC. There's some information about how they trained the model. If you have an interest in that, please definitely take a look. And let's see, I'll tell you, they used the

Ben Courchia, MD (06:57.989) No, walk us through that.

Ben Courchia, MD (07:03.49) No, I'm kidding. I'm kidding. I'm kidding. I'm kidding.

Daphna Yasova Barbeau, MD (07:06.339) You know more about this model training than I do. But they had it learn, and I think it learned pretty well. I'll tell you about how it learned. So across the study period, they screened 211 infants. They had 50 infants who were ineligible, and then 25 they couldn't get consent. 40 parents declined. So they enrolled just under 100 infants. The range of corrected gestational ages and weights

at any time of measurement were 25 to 36 weeks and 600 grams to 3,380 grams. Of note, they did have some parents decline, but they note that parental willingness to consent was pretty high. Most parents had very few perceived concerns with respect to the BOS apparatus. They didn't observe any patient harm. There were no parental or nursing complaints received.

And voluntary withdrawal was available at any time and routinely offered at the time of consent, but wasn't requested by any of the parents or clinicians. They found that the BOS measurements were noted to be very reliable and very consistent in a number of different ways. And the figures I think are pretty cool. So they're basically like these, if you're a Floridian, you know what a spaghetti plot is for hurricanes. So this looks like a

Ben Courchia, MD (08:28.232) The stress.

Daphna Yasova Barbeau, MD (08:29.104) This looks like a spaghetti plot. But basically, I'll tell you, on the vertical axis you have this normalized reflectance, so a measurement of reflectance, and then across the horizontal axis you have the wavelength. So you can see that almost all of the measurements across the wavelengths have this predictable pattern.

The pattern is kind of random, but different wavelengths provide different patterns. And so I'll just describe the rest of this figure. So they looked at the BOS reproducibility. So this first figure is just one neonate without NEC over the course of a 10 week period, looking at these wavelengths. And you can see even 10 weeks apart that the measurements are pretty similar to one another at each

wavelength. And then this second graph is looking at reproducibility and variability among neonates without NEC. And they used this weight of 1.4 to 1.5 kilos. And every line on the graph is an individual neonate, but you can tell that they follow a basically identical pattern at every wavelength. Yeah. So it's pretty cool that they were able to show that.

Ben Courchia, MD (09:49.762) Kinda crazy, huh?

Ben Courchia, MD (09:55.085) Yeah, usually you see a lot of variability, but to be honest with you, if you zoom out on the PDF, you may not even detect that it's individual units. It looks like one single line, it's so homogeneous.

Daphna Yasova Barbeau, MD (09:57.703) Right.

Daphna Yasova Barbeau, MD (10:01.734) That's multiple lines, right?

Yeah. And so they wanted to show, how does the variability change for one baby? What does the variability look like among other babies? And then they wanted to describe how is BOS able to distinguish these underlying abdominal changes from, say, the thigh, like other soft tissues, to look at the difference in the viscera basically.

So then they direct you to figure three, which shows you the difference in BOS reflectance across different wavelengths in different tissues. So let's say in this first graph, in pink, they're showing us what does the abdomen look like and what does the thigh look like. And to be perfectly honest, they don't look that different to me.

But what you can tell is that there is just this predictable pattern in both tissues, depending on the reflectance at each individual wavelength. So they wanted to look at that, like I said, the difference between readings taken from the abdomen versus the lateral thigh.

That was the first stage of looking at differences in BOS. Then they wanted to look at, if you had normal versus abnormal bowel, what does that look like? And I'll describe that next figure to you. But 10 of the infants developed NEC over the course of the study, and an additional four infants had a spontaneous intestinal perforation, or they say, or other intraabdominal process. They were characterized by abdominal distension and pneumatosis, plus bloody stools, and

Daphna Yasova Barbeau, MD (12:01.044) portal venous gas was observed in some. In all of the cases, they had a routine procedure for managing those babies. They held the feeds immediately, initiated broad spectrum antibiotics. And two out of the 10 infants did require laparotomy. And so they were, I think, pleased that they were able to demonstrate detectable differences on spectroscopy between infants with severe NEC compared to a time point when the same infant did not have NEC.

So basically, they show us a graph of three different babies, looking at their time point across wavelengths and looking at the reflectance. So it looks like when they're unaffected, they highlighted in red, it's a much higher reflectance, especially between the wavelengths of about, I don't know, 780

and 1650. And you can see the severe NEC cases, they're outlined in black, have a big drop in the reflectance. I'd say like a five to ten percent drop in the reflectance. And that's where you can see there's some changes across the wavelength spectrum. In general, the severe NEC cases have a lower reflectance, but that's not true for all the cases.

But especially across the wavelengths in this 780 to 1650 range, you can see the difference between babies who have NEC and babies who don't have NEC. So, machine learning. So again, briefly, they used the model to project these features,

looking at NEC status with a partial least squares discriminant analysis. And the area under the curve was shown to be very high, indicating good overall separation between NEC and non-NEC scans. I mean, the area under the curve here in the figure is almost 100%. It looks very good. And then you can see they list the sensitivity as a hundred percent, specificity of

Daphna Yasova Barbeau, MD (14:19.172) 94%, a negative predictive value of 100%, which is pretty valuable. And then they do discuss this in the discussion, but the positive predictive value is only about 24%. But you do have those impressive values in the sensitivity, specificity, and negative predictive values. So

they describe that overall the model shows outstanding sensitivity, specificity, and negative predictive value. And they felt that, quote, the presence of detectable changes in light reflectance in these premature infants with moderate or severe NEC suggests that BOS shows promise as a useful tool in this vulnerable population. So their hope is that there are

basically early color changes to the affected bowel when we have the onset of ischemia. So it's not just detecting oxygenation, but it's actually detecting color change in the underlying bowel. And because babies' skin is thin, they don't have a lot of overlying fat or muscle that will interfere with being able to detect the light reflectance.

Ben Courchia, MD (15:25.4) Mm-hmm.

Daphna Yasova Barbeau, MD (15:35.049) And they do talk about the difference between NIRS and BOS, which we discussed, and they feel that potentially BOS will give us more information than NIRS alone. So they feel that BOS is safe and feasible. It gives rapid point-of-care measurements that you can look at over the course of time. Obviously, this is a small study, but it is exciting,

especially as you told us yesterday, with the loss of probiotics and an increase in NEC cases, that we might have another tool that is non-invasive, potentially not that expensive. I don't know, I'm saying that without any knowledge of what this costs, but it doesn't seem that expensive. It seems like it's this little handheld thing. Cost of BOS machine.

Ben Courchia, MD (16:09.815) Mm-hmm.

Ben Courchia, MD (16:19.407) You have no clue. What? Have you seen that machine? It looks big.

No, no, no, no, no. It looks bigger. I looked it up. Look at that. It looks big. It's

Daphna Yasova Barbeau, MD (16:35.26) Okay. It's not, let me see.

Ben Courchia, MD (16:37.977) Let me show you. Let me show you. Hold on. Allow window. I'm gonna show you. This is it. Hold on. Share. This is it. This is the machine.

Daphna Yasova Barbeau, MD (16:46.62) Listen, a spectroscopy machine. This is a different one.

Ben Courchia, MD (16:52.845) You see it? So the handheld part is small. It can, okay, I'm gonna give, I don't know who makes this machine. I have no vested interest in this machine, but I'm gonna make the best marketing pitch for them. It looks like the Ghostbusters. That's very cool.

Daphna Yasova Barbeau, MD (16:53.98) Yeah, I see it.

Yeah.

Daphna Yasova Barbeau, MD (17:09.436) That's fair. But that's not bad. This is like the size of a projector machine and a handheld ultrasound probe. Yeah, that's right. Just check every baby, just one to one. I love that. All right. Thoughts?

Ben Courchia, MD (17:14.957) Yeah. It looks like the Ghostbusters. You put it on your back and you come with this thing and you'd be like choo choo choo.

My thoughts are that it's not ready for prime time, but it's very exciting. I find that what you were describing earlier, specifically when we look at these cases of NEC, it's not a clear departure from the curve. It's not like there's crossing, it's just a dampening of the curve. But I feel like it's such an early paper that

Daphna Yasova Barbeau, MD (17:31.688) Yeah.

Daphna Yasova Barbeau, MD (17:43.09) Yeah.

Ben Courchia, MD (17:51.999) you know that more stuff is gonna come down the pike, trying to look at specific, maybe creating even a nomogram, who knows? That initial, the little table, the four squares table that we do in genetics, my god, the Punnett square almost, with the true positives, that's very promising as well. And then I think that you were discussing how this really looks at the bowel color itself.

Daphna Yasova Barbeau, MD (18:00.242) Mm-hmm.

Daphna Yasova Barbeau, MD (18:09.404) Mm-hmm.

Daphna Yasova Barbeau, MD (18:20.231) Mm-hmm.

Ben Courchia, MD (18:20.907) Obviously, where I really want this to do well is for the medical cases of NEC. Obviously the babies who have dead bowel or who have a perforation, those are not usually the ones where we struggle the most when it comes to diagnosing, but there are gonna be all these in between, where maybe the signal can be identified and maybe this can pick up a bunch of stuff in a more sensitive manner. That would be very exciting. I'm looking at that with excitement.

Daphna Yasova Barbeau, MD (18:27.985) Mm-hmm.

Daphna Yasova Barbeau, MD (18:48.284) Yeah, and I don't know. Will it look like nomograms, or will the babies act as their own controls, and seeing a change for the specific babies? I don't know. They'll have to let us know. Yeah.

Ben Courchia, MD (18:58.147) We'll see. But it's definitely exciting. And I think that we always lament the lack of innovation or the lack of new techniques. This is innovative. This is not repurposing anything that we currently are using. This is definitely new. So you heard it first on the Incubator. Well, thank you very much, buddy. It has been a long journal club week. Thank you. We'll be back tomorrow with a special guest. We're gonna skip the news tomorrow and have

Daphna Yasova Barbeau, MD (19:03.933) Mm-hmm.

Mm.

Daphna Yasova Barbeau, MD (19:11.112) For sure. Cool. That's right.

Daphna Yasova Barbeau, MD (19:20.648) Mm-hmm.

Ben Courchia, MD (19:26.891) an extended journal club session, so stay tuned for that, and see you tomorrow.

Daphna Yasova Barbeau, MD (19:31.433) Sounds good. Bye everyone.