Neonatal brain imaging

Introduction

Optical tomography of the neonatal brain is being developed to help doctors diagnose and treat newborn babies suffering from hypoxic-ischaemic brain injury. This occurs when breathing difficulties or other problems prevent the baby from receiving sufficient amounts of oxygenated blood during birth. This is a major cause of permanent disability in very preterm infants who survive after neonatal intensive care. The objective of our work is to acquire three-dimensional (3D) images of the infant brain which reveal the variation in blood volume and tissue oxygenation. Currently there is no alternative method which can be used safely and continuously on infants in intensive care.

monstirimages


Figure 1: A custom-built helmet for optical
tomography of the neonatal brain.

monstirimages


Figure 2: Reference phantom inserted within a neonatal imaging helmet.

To acquire measurements of light transmitted between multiple locations on the surface of a newborn infant head, our initial imaging studies have utilised interfaces custom-built for each individual infant in the form of a plastic, foam-lined helmet. Each helmet (see figure 1) fits over the back and top of the head. The outer shell of the helmet is constructed in two halves from low-temperature thermoplastic, and is lined with a soft NIR-absorbing foam. The lower half, which fits beneath the rear hemisphere of the head, supports the weight of the head and is held a few centimetres above the cot by a plastic frame. The combined source/detector fibre bundles are attached to the helmet via small sockets mounted on the thermoplastic shell, and their positions are measured with a 3D digitizer. However, since the helmet is necessarily deformable, the recorded positions are inevitably somewhat approximate. To reduce the sensitivity of the image reconstruction algorithm to the uncertainty in the true source/detector positions, we acquire a reference measurement using the same helmet and an object with precisely known optical properties. This object consists of a balloon filled with a solution of intralipid and near-infrared dye, with optical properties similar to those of brain tissue. Figure 2 shows the reference phantom placed within a helmet.

monstirimages


Figure 3: Moving MONSTIR to the
neonatal clinic in October 2001.

monstirimages


Figure 4: A helmet attached to the head of
a 10-day-old, 30 weeks gestation infant.

Intraventricular haemorrhage

MONSTIR was transported across the street to the UCL neonatal intensive care unit for the first time in October 2001 (figure 3). Initial studies on two unventilated premature infants enabled the helmet design to be evaluated and modified, and established that it was sufficiently comfortable to be worn continuously for several hours without interfering with the normal handling of the infant (see figure 4). For a third study, the reference phantom was employed for the first time, enabling us to attempt an image reconstruction using difference data. The study involved a 5-week-old male infant, born after 30 weeks gestation, who had suffered from a perinatal haemorrhage. An ultrasound scan (see figure 5) revealed a large left intraventricular haemorrhage and a much smaller right intraventricular haemorrhage. The helmet contained 31 sources and detectors distributed over the back and top of the head. Data acquisition proceeded automatically as each source was illuminated for 15 seconds, and data were recorded by each detector simultaneously. Each scan required about 9 minutes. Calibration measurements were also acquired, which involved detecting light back-reflected at the skin surface while the source illumination was heavily attenuated. Two full sets of data were acquired on the infant, followed by an identical measurement on the reference phantom inflated to exactly fill the helmet after the infant had been carefully removed.

In order to generate 3D images of the internal optical properties, TOAST requires a FEM model of the infant head with a realistic geometry. To provide a suitable mesh, we first acquired a 3D CT-scan of a realistic doll's head, from which we generated a surface mesh. Appropriate software was then used to apply a non-linear warp to the surface mesh in order to fit it to the measured locations of the sources and detectors on the helmet. Finally, the resulting surface was used to construct a volume mesh, containing 17559 second-order tetrahedral elements having a total of 26814 nodes. Three-dimensional images of the neonatal head were generated using differences between the mean flight times measured for the head and the corresponding values for the intralipid filled balloon. TOAST performed 25 iterations using the data recorded at each wavelength, starting from a homogeneous estimate with properties corresponding to those of the reference phantom. Each iteration required 25 minutes on a 2.2 GHz PC with 2 Gb of RAM.

monstirimages


Figure 5: An ultrasound scan of an infant brain with a large
haemorrhage in the left ventricle.

monstirimages


Figure 6: Transverse, coronal, and sagittal views across the
3D absorption image of the infant, acquired at 780 nm.

Figure 6 shows transverse, coronal, and sagittal views across the 3D reconstruction of the absorption coefficient at a wavelength of 780 nm. A 3D image of scatter coefficient was generated simultaneously (see Hebden et al [1] for more details), and similar images of both coefficients were acquired at 815 nm. The sagittal view corresponds to the midplane, and the coronal and transverse views are centred on the expected location of the cerebral ventricles. The coronal image above clearly exhibits greater absorption on the left side of the brain, consistent with the larger intraventricular haemorrhage. The transverse and sagittal images are dominated by a feature at the rear of the head, possibly due to the proximity of the sagittal sinus, since it represents a dominant source of venous blood close to the surface. A full description of the results and the imaging method has been published [1,2].

Imaging changes in blood volume and oxygenation

In August 2002 a study was performed on a four-day-old female term infant who, as a result of severe global hypoxic-ischaemic insult following uterine rupture, required mechanical ventilation and sedation. A customised helmet was constructed to support 29 source/detector pairs around her head (figure 7). Over a three hour period a series of eight optical imaging scans were performed between incremental changes to the ventilator settings, which facilitated changes to the partial pressures of carbon dioxide (PaCO2) and oxygen (PaO2) which were monitored throughout the study. Each scan involved illuminating each source for either 5 or 15 seconds while TPSFs were recorded by each detector simultaneously. Figure 8 shows the average photon detection rate (photons per second per active detector) calculated for each set of data and for both source wavelengths.

monstirimages


Figure 7: Imaging a ventilated infant
in neonatal clinic in August 2002.

monstirimages


Figure 8: Average detector photon count rates
during changes to ventilator settings.

A 3D digitizer was used to measure the locations of the sources/detectors on the corresponding helmet, and FEM meshes of each infant head was generated by warping a generic head surface mesh to the measured co-ordinates of the sources and detectors. In each case, values of integrated intensity and mean flight time were calculated from each TPSF after appropriate calibration. Image reconstruction was performed using selected pairs of data sets at both wavelengths. TOAST first calculates the differences between the values of mean flight time and log intensity for each active source-detector combination. The algorithm then employs an iterative procedure to generate 3D images of absorption and scatter which yield forward data as close as possible to the measured difference data added to the forward prediction of an initial estimate. For each reconstruction, TOAST performed 20 iterations, although no noticeable improvement in image quality was observed after about the tenth iteration. Each iteration typically required about 7 minutes on a 2.2 GHz PC with 2 Gb of RAM.

monstirimages

Figure 9: Slices across the 3D images of the infant head, representing differences in absorption at 815 nm due to increased PaCO2.

monstirimages

Figure 10: Sagittal slices representing differences in absorption at 815 nm and 780 nm due to decreased PaO2.

Figure 9 shows slices across the 3D absorption image of the ventilated infant brain corresponding to the difference between an initial baseline measurement and data acquired at higher PaCO2 at a wavelength of 815 nm. These confirm the large global absorption increase due to increased CBV, with evidence of division into left and right hemispheres, and possibly of the blood-free ventricles within the centre. Figure 10 shows sagittal slices from the absorption images acquired at both wavelengths corresponding to a decrease in PaO2 at constant PaCO2. As expected, a modest increase in absorption is revealed at 780 nm, while at 815 nm we observe slight decreases in absorption within localised regions. Scatter difference images (not shown) were also generated but small displayed features were probably due to imperfect separation between scatter and absorption. A more thorough description of this clinical measurement and of the results obtained is available in a recent publication [3].

Whole-brain imaging of evoked response

Optical tomography studies have been performed on a cohort of six pre-term babies in order to identify regions of the brain activated during passive motor evoked response [4]. Data were acquired during bilateral passive arm movement, and 3D images of the entire infant head were reconstructed. MONSTIR recorded data from 32 detector positions simultaneously while the head was illuminated at a single source position for ten seconds during activation, and then for another ten seconds during a period of rest. This was repeated for 12 source positions. A linear reconstruction method was used to generate images which represent the changes in absorption and scatter occuring in the head due to the activation.

monstirimages


Figure 11: Sagittal slices across 3D image of absorption change in infant brain due to passive movement of left arm.

monstirimages


Figure 12: Sagittal slices across 3D image of absorption change in infant brain due to passive movement of right arm.

Figure 11 shows the change in absorption within the brain of a 33-week-old male infant during passive movement of the left arm, and figure 12 shows the corresponding change for a 34-week-old female infant during passive movement of the right arm. The dominant feature in both cases is an increase within the contralateral cerebral hemisphere near the expected position of the motor cortex. Figure 12 also shows a corresponding decrease within the anterior region, possibly resulting from a transitory transfer of blood towards the motor cortex. The source wavelength in each case is 780 nm. By combining images recorded at 780 nm and 815 nm, we have been able to generate images of changes in concentrations of oxy- and deoxy-haemoglobin, and of total haemoglobin [4].

References

  1. Hebden, JC, Gibson, A, Yusof, R, Everdell, N, Hillman, EMC, Delpy, DT, Arridge, SR, Austin, T, Meek, JH, and Wyatt, JS (2002): Three-dimensional optical tomography of the premature infant brain, Physics in Medicine and Biology 47, 4155-4166. Download PDF file.
  2. Austin, T, Hebden, JC, Gibson, AP, Branco, G, Yusof, R, Arridge, SR, Meek, JH, Delpy, DT, and Wyatt, JS (2006): Three-dimensional optical imaging of blood volume and oxygenation in the preterm brain, Neuroimage 31, 1426-1433. Download PDF file.
  3. Hebden, JC, Gibson, A, Austin, T, Yusof, R, Everdell, N, Delpy, DT, Arridge, SR, Meek, JH, and Wyatt, JS (2004): Imaging changes in blood volume and oxygenation in the newborn infant brain using three-dimensional optical tomography, Physics in Medicine and Biology 49, 1117-1130. Download PDF file.
  4. Gibson, AP, Austin, T, Everdell, NL, Schweiger, M, Arridge, SR, Meek, JH, Wyatt, JS, Delpy, DT, and Hebden, JC (2006): Three-dimensional whole-head optical tomography of passive motor evoked responses in the neonate, Neuroimage 30, 521-528. Download PDF file.