Five healthy controls (two males and three females, 2–14 years) were measured for normal values of the brain. For healthy controls, both absorption and reduced scattering coefficients were similar between the laterally symmetrical positions (FP1 vs. FP2, F3 vs. F4, and C3 vs. C4. All p values are >0.05.). Thus, the quantified lateralization factors L a and L s were close to zero (all values <0.06). In contrast, the L a and L s values from the ECMO patients are largely variable. Figure 3, A and B shows the cumulative distribution function of L a and L s at the group level, respectively. In each figure, all of the values from healthy controls are within a narrow range of 0–0.06. The dash line indicates three standard deviations (single-sided normal distribution) of data from the control group, which constrains 99.7% probability of data variation from the normal brains. Thus, it was used as a threshold to identify the abnormalities among ECMO patients. As a result, 24% of L a values and 16% of L s values from the ECMO patients exceeded this threshold. Results from individual ECMO patients are described below.
Patient 1 was a 1 year old female who was a premature infant born at 28 weeks of gestation and placed on VV-ECMO for ARDS secondary to H1N1 pneumonia. She stayed on VV-ECMO for 12 days and was then weaned off ECMO support. Very thick hair resulted in NIRS measurements being feasible only at the FP1/FP2 points on three dates (days 1, 2, and 4). No regional abnormality was seen in either absorption or scattering although the post-ECMO MRI demonstrated volume loss of the white matter in bilateral thalami and cerebellar hemispheres (Figure 5, top-left panel), which were thought to be from underlying prematurity.
Patient 2 was a 9 month old female who was placed on VV-ECMO for pulmonary hypertensive crisis. During the ECMO run, she was converted to VA-ECMO due to hemodynamic instability. She stayed on VV/VA-ECMO for totally 6 days and was then weaned off. The NIRS measurements were conducted on three different dates (days 1, 2, and 4) during ECMO. Large variations in absorption coefficient were seen over time, especially at F3/F4, which therefore led to large lateralization difference (L a > 0.1) between F3 vs. F4 and C3 vs. C4. These variations corresponded to the areas of large subdural fluid accumulation demonstrated by MRI, which were indicated pre-ECMO and significantly worsened post-ECMO, as shown in Figure 4C. For this patient, the reduced scattering coefficient was more stable over time and more consistent laterally.
Patient 3 was a 3 day old full-term female who was placed on VV-ECMO for PPHN and meconium aspiration. She stayed on VV-ECMO for 4 days and was then weaned off. For this patient, only the FP1/FP2 points were measured by NIRS on two dates (days 1 and 3) due to the presence of thick hair at all the other positions. On day 1, both absorption and reduced scattering coefficients showed significant reduction on the left cerebral hemisphere. On day 3, both coefficients returned to normal. The early abnormality seen in the NIRS corresponded to abnormal signal intensity within the white matter of the same region shown in the post-ECMO MRI (Figure 5, top-right panel).
Patient 4 was a 14 year old female who was placed on VV-ECMO for ARDS secondary to H1N1, as well as methicillin-resistant Staphylococcus aureus(MRSA) pneumonias. During her ECMO run, she was converted to VA-ECMO due to hemodynamic instability. She stayed on VV/VA-ECMO for 20 days and was then weaned off. For this patient, the FP1/FP2 and F3/F4 points were measured by NIRS (the C3/C4 points were blocked by cannulas) on three successive dates (days 1–3). All results were normal. However, the post-ECMO MRI did have some evidence of mild atrophy and small area of hemorrhage in the right centrum semiovale (Figure 5, bottom-left panel).
Patient 5 was a 6 month old female who was placed on VV-ECMO for ARDS secondary to H1N1. She was on VV-ECMO for 12 days before her death. Near-infrared spectroscopy measurements were conducted on two successive dates (days 1 and 2) in the late course of her ECMO run. While the absorption coefficient was relatively stable over time, a decrease in reduced scattering coefficient from day 1 to day 2 was seen in all six positions, especially in F3 and C3. As a result, the corresponding lateralization difference between F3 vs. F4 and C3 vs. C4 were very large (L s > 0.2) in day 2. Patient died from right atrial perforation from cannula malpositioning during her ECMO run.
Patient 6 was a 16 year old female who was placed on VV-ECMO, with a femoral drainage cannula, for ARDS secondary to respiratory syncytial virus (RSV) septic shock. She stayed on VV-ECMO for 26 days and was then weaned off. The NIRS measurements were conducted on three dates (days 1, 4, and 5) and all of the results were normal. There was no available post-ECMO imaging as the patient was diagnosed of incurable cancer during her ECMO therapy.
Patient 7 was a 2 day old full-term male who was placed on VA-ECMO for PPHN and meconium aspiration. He was on VA-ECMO for 3 days and then weaned off. For this patient, the FP1/FP2 and F3/F4 points were measured by NIRS (the C3/C4 points were blocked by cannulas) on three successive dates (days 1–3). Significant reduction in reduced scattering coefficient was seen in the left cerebral hemisphere (Fp1 on day 1 and F3 on day 3), which corresponded to the area of restricted diffusion involving the superior cerebellar peduncles shown in MRI. This indicates that ischemic injury to the involved area could lead to decrease in white and grey matter within the central mid brain (Figure 5, bottom-right panel).
At the group level, it was observed that lower-than-normal values of reduced scattering coefficient (µ s′) in one side of the brain were consistently associated with large lateralization difference (L s), such as patient 5 above, whereas the relationship between absorption coefficient (µ a) and lateralization difference (L a) was not apparent. Thus, linear regressions were further conducted to assess the relationships between L s and µ s′ as well as between L a and µ a, and the results are shown in Figure 6, A and B, respectively. Because each nonzero lateralization factor corresponded to a higher µ a or µ s′ value on one side of the brain and a lower value on the other side of the brain, each of the L a − µ a or L s − µ s′ relationship was analyzed based on its corresponding higher-value brain side and lower-value brain side separately. These analyses demonstrated that the µ a values had little correlation with L a on both higher-value brain side (R 2 = 0.05, p = 0.18) and lower-value brain side (R 2 = 0.03, p = 0.32). Similarly, µ s′ on the higher-value brain side also had little correlation with L s (R 2 = 0.07, p = 0.12). However, µ s′ on the lower-value brain side had a significant descending correlation with L s (R 2 = 0.35, p < 0.01). Therefore, it appears that the large lateral difference in scattering was caused by a reduction of scattering in one side of the brain.
The current study assessed intracranial abnormalities in neonatal and pediatric patients on ECMO support by measuring the regional light absorption and scattering with frequency-domain NIRS technology. The precision of the NIRS measurements enabled identification of regional abnormalities in both absorption and scattering based on comparisons between healthy controls and ECMO patients, as well as between the laterally symmetrical positions (FP1 vs. FP2, F3 vs. F4, and C3 vs. C4 in Figure 2). The results from this study indicate a potential to use the frequency-domain NIRS for detection of intracranial complications during ECMO therapy.
ECMO is associated with a relatively high incidence of substantial complications, of which intracranial injuries are the most important and most devastating.9 These injuries consist of hemorrhagic and nonhemorrhagic, ischemic lesions. The existence of intracranial injuries and the classifications of severity are the major predictors for neurodevelopmental outcomes in survivors. In general, head ultrasound is used for detection of intracranial abnormalities in ECMO-treated newborns and infants with open fontanels. It is sensitive in the evaluation of large intracranial hemorrhage,22 but less useful for the detection of ischemic lesions.
Several studies have demonstrated that CT and MRI scans can identify intracranial lesions that earlier were not seen with head ultrasound. For instance, a recent study by Rollins et al.23 studied 50 neonates with MRI done after ECMO, before discharge, and compared the results with head ultrasound completed during ECMO. Head ultrasound was abnormal in 24% neonates during ECMO, whereas post-ECMO MRI was abnormal in 62% neonates. Most of the lesions unrecognized on head ultrasound were nonhemorrhagic. While MRI appears to be ideal for post-ECMO evaluation, its application during ECMO therapy is not possible. CT imaging is limited by the difficulty in transport of these patients to the radiology suite and the potential risks, such as accidental decannulation and death. Thus, we currently lack a reliable tool for early detection of intracranial injuries during ECMO therapy, especially for the patients with closed fontanels where head ultrasound cannot be used.
Near-infrared spectroscopy has been widely accepted as a noninvasive, portable technology in critical care unit that has demonstrated sensitivity to changes in cerebral tissue oxygenation.14,15 Compared with head ultrasound, NIRS can be used for all age groups because the near-infrared light can easily pass through the skull. A main drawback of NIRS is its limited penetration depth. The depth sensitivity of NIRS increases with the source-detector distance. With an achievable source-detector distance of 3 to 5 cm, NIRS has sufficient sensitivity to the cortical sulci and superficial white matter.13 On the other hand, a study by Bulas et al.22 found that hemorrhages in ECMO-treated newborns are mainly within the brain parenchyma. Thus, we suggest NIRS could still be a complementary tool to the existing technologies for bedside evaluation during ECMO therapy.
The frequency-domain NIRS used in this study determined absolute absorption and reduced scattering coefficients of cerebral tissues based on the incremental changes of light attenuation and phase shift at four source-detector distances. In this way, the light attention and phase shift caused by the extracerebral tissues (i.e., the scalp and skull) were represented as the base values at the shortest source-detector distance, and therefore had minimal influence on the cerebral readings. This mechanism limited the measurement variations in both absorption and scattering, which allowed us to investigate their abnormal changes by comparing readings between healthy controls and ECMO patients, as well as between the laterally symmetrical positions. A main finding in this study is that the identified abnormalities in scattering among ECMO patients appear to be associated with lower-than-normal values of reduced scattering coefficient seen in regional areas of the brain. Because light scattering originates from the intracellular structures (such as nuclei and mitochondria),16 a reduction in scattering primarily reflects loss or decreased density of the brain matter. Loss of brain volume can be because of several different factors, such as resolving hemorrhage, hypoxic ischemic lesions, and injury secondary to stroke, all of which can occur in patients undergoing ECMO therapy. Furthermore, we have also seen the consistent changes in the scattering corresponding to changes in post-ECMO neuroimaging. However, this correspondence was seen only if the injury was in the parietal or frontal regions of the brain. The NIRS measurement was not sensitive enough to pick up changes in the water shed areas of the brain, such as hippocampi and the deep cerebral and cerebellar nuclei.
Because none of the ECMO patients in this study developed significant intracranial bleeding during the therapy, we did not see any consistent changes in the absorption pattern. Large intracranial bleeding would certainly cause significant changes of blood volume in the area of the brain that is involved, which would lead to abnormalities in the absorption pattern. However, further studies on larger groups of patients during ECMO are needed for confirmation. Furthermore, frequency-domain NIRS was done during the course of the ECMO run when patients were much sicker and had significant disturbances in their cerebrovascular functions.24 The neuroimaging done is much later in the course of their hospitalization when they are less sick and stable. This can be one of the significant reasons why the frequency-domain NIRS changes did not always correlate with the changes seen on neuroimaging.
Another limitation in this study was the light absorption by the hair, for which data from several cranial points (such as F3, F4, C3, and C4) could not be collected among subjects with thick hair. The hair absorption is a common issue in NIRS. This can be improved by better probe design, for example, letting the optical fibers have short tips (1–2 mm) out of the probe to pass through the hair. Furthermore, recently, we have reported the use of brush optodes to improve the optical contact on subjects with dense hair.25 The brush optode is applicable for all kinds of NIRS probes, which can be regarded as the ultimate solution of this issue.
Regional abnormalities in both absorption and scattering were identified by frequency-domain NIRS among pediatric patients during ECMO therapy. The scattering abnormalities appear to be associated with lower-than-normal µ s′ values in regional areas of the brain, which primarily reflects loss or decreased density of the brain matter. These results suggest that the frequency-domain NIRS potentially can be a complimentary bedside tool to help detect cerebral abnormalities for patients on ECMO. Future studies are needed to correlate these abnormalities with long-term outcomes.
1. Lequier L. Extracorporeal life support in pediatric and neonatal critical care: A review. J Intensive Care Med 2004.19: 243–258.
3. UK Collaborative ECMO Trial Group: UK collaborative randomised trial of neonatal extracorporeal membrane oxygenation
. Lancet 1996.348: 75–82.
4. Polito A, Barrett CS, Wypij D, et al. Neurologic complications in neonates supported with extracorporeal membrane oxygenation
. An analysis of ELSO registry data. Intensive Care Med 2013.39: 1594–1601.
5. Hofkosh D, Thompson AE, Nozza RJ, Kemp SS, Bowen A, Feldman HM. Ten years of extracorporeal membrane oxygenation
: Neurodevelopmental outcome. Pediatrics 1991.87: 549–555.
6. Glass P, Wagner AE, Papero PH, et al. Neurodevelopmental status at age five years of neonates treated with extracorporeal membrane oxygenation
. J Pediatr 1995.127: 447–457.
7. Kasirajan V, Smedira NG, McCarthy JF, Casselman F, Boparai N, McCarthy PM. Risk factors for intracranial hemorrhage in adults on extracorporeal membrane oxygenation
. Eur J Cardiothorac Surg 1999.15: 508–514.
8. Glass P, Bulas D. Neonatal ECMO: Neuroimaging and neurodevelopmental outcome. Semin Perinatol 2005.29: 58–65.
9. van Heijst AF, de Mol AC, Ijsselstijn H. ECMO in neonates: Neuroimaging findings and outcome. Semin Perinatol 2014.38: 104–113.
10. Lidegran MK, Mosskin M, Ringertz HG, Frenckner BP, Lindén VB. Cranial CT for diagnosis of intracranial complications
in adult and pediatric patients during ECMO: Clinical benefits in diagnosis and treatment. Acad Radiol 2007.14: 62–71.
11. Ferrari M, Quaresima V. A brief review on the history of human functional near-infrared spectroscopy (fNIRS) development and fields of application. Neuroimage 2012.63: 921–935.
12. Cui X, Bray S, Reiss AL. Functional near infrared spectroscopy (NIRS) signal improvement based on negative correlation between oxygenated and deoxygenated hemoglobin dynamics. Neuroimage 2010.49: 3039–3046.
13. Dehghani H, White BR, Zeff BW, Tizzard A, Culver JP. Depth sensitivity and image reconstruction analysis of dense imaging arrays for mapping brain function with diffuse optical tomography. Appl Opt 2009.48: D137–D143.
14. Fenik JC, Rais-Bahrami K. Neonatal cerebral oximetry monitoring during ECMO cannulation. J Perinatol 2009.29: 376–381.
15. Wong JK, Smith TN, Pitcher HT, Hirose H, Cavarocchi NC. Cerebral and lower limb near-infrared spectroscopy in adults on extracorporeal membrane oxygenation
. Artif Organs 2012.36: 659–667.
16. Wang LV, Wu H-I. Biomedical Optics: Principles and Imaging. 2007.New Jersey, Wiley-Interscience.
17. Donofrio MT, Massaro AN. Impact of congenital heart disease on brain development and neurodevelopmental outcome. Int J Pediatr 2010.359390.
18. Miller SP, McQuillen PS, Hamrick S, et al. Abnormal brain development in newborns with congenital heart disease. N Engl J Med 2007.357: 1928–1938.
19. McQuillen PS, Barkovich AJ, Hamrick SE, et al. Temporal and anatomic risk profile of brain injury with neonatal repair of congenital heart defects. Stroke 2007.38(2 suppl): 736–741.
20. Arri SJ, Muehlemann T, Biallas M, Bucher HU, Wolf M. Precision of cerebral oxygenation and hemoglobin concentration measurements in neonates measured by near-infrared spectroscopy. J Biomed Opt 2011.16: 047005.
22. Bulas DI, Taylor GA, O’Donnell RM, Short BL, Fitz CR, Vezina G. Intracranial abnormalities in infants treated with extracorporeal membrane oxygenation
: update on sonographic and CT findings. AJNR Am J Neuroradiol 1996.17: 287–294.
23. Rollins MD, Yoder BA, Moore KR, et al. Utility of neuroradiographic imaging in predicting outcomes after neonatal extracorporeal membrane oxygenation
. J Pediatr Surg 2012.47: 76–80.
24. Short BL. The effect of extracorporeal life support on the brain: A focus on ECMO. Semin Perinatol 2005.29: 45–50.
25. Khan B, Wildey C, Francis R, et al. Improving optical contact for functional near-infrared brain spectroscopy and imaging with brush optodes. Biomed Opt Express 2012.3: 878–898.
Keywords:Copyright © 2017 by the American Society for Artificial Internal Organs
extracorporeal membrane oxygenation; frequency-domain near-infrared spectroscopy; intracranial complications; absorption; scattering