Development and characterization of transfontanelle photoacoustic imaging system for detection of intracranial hemorrhages and measurement of brain oxygenation: Ex-vivo

Abstract

We have developed and optimized an imaging system to study and improve the detection of brain hemorrhage and to quantify oxygenation. Since this system is intended to be used for brain imaging in neonates through the skull opening, i.e., fontanelle, we called it, Transfontanelle Photoacoustic Imaging (TFPAI) system. The system is optimized in terms of optical and acoustic designs, thermal safety, and mechanical stability. The lower limit of quantification of TFPAI to detect the location of hemorrhage and its size is evaluated using in-vitro and ex-vivo experiments. The capability of TFPAI in measuring the tissue oxygenation and detection of vasogenic edema due to brain blood barrier disruption are demonstrated. The results obtained from our experimental evaluations strongly suggest the potential utility of TFPAI, as a portable imaging modality in the neonatal intensive care unit. Confirmation of these findings in-vivo could facilitate the translation of this promising technology to the clinic.

Keywords: Hemorrhage; Neonates; Oxygenation; Photoacoustic imaging; Transfontanelle.

Fig. 1

Different types and locations of intracranial hemorrhages, ICH (epidural, subdural, subarachnoid, intracerebral, intraventricular). IVH: intraventricular hemorrhage; SSS: superior sagittal sinus; Grade IV: extension of IVH into the intracerebral hemorrhage.

Fig. 2

TFPAI system components. (a) Optical setup of laser light coupling to fiber bundle with (i) laser light coupling optics, (ii) laser light, (iii) laser system head, (iv) optical breadboard, and (v) optical post and post holder. (b) Laser light coupling optics cut section view with (i) fiber optic cable bundle, (ii) convex lens, (iii) diffuser, (iv) spacer, (v) parabolic reflector, (vi) thin silver coating, (vii) laser light, and (viii) horizontal cage system. (c) TFPAI probe including (i) US data cable, (ii) fiber optic cable bundle, (iii) flexible strap, (iv) linear array transducer, (v) fiber optic cable housing, (vi) fiber optic bundle housing, (vii) fiber optic cable bundle, (viii) main body, (ix) inner fiber optic cable holder, (x) outer fiber optic cable holder, (xi) soft tip, and (xii) Aqualene® coupler. (d) Rendering of TFPAI probe in use.

Fig. 3

Simulation studies for optimum light delivery. (a) (i)Top view: an US transducer is placed at the fontanelle center (darker grey outline) and optical fibers (yellow circles) located at the sides act as light sources. Fontanelle is considered as a 20 × 20 mm square (black box). Distance from transducer wall and the first row is L, (ii) Side view: the distance between fibers and tissue surface is H and bending angle of fibers is θ, (iii) fiber diameter, D, fiber positioning in a square (left) and honeycomb (right) configurations. The separation between adjacent fibers is 0.5 mm in both cases, (iv) different fiber arrangements: one row at one side (top left), one row on each side (top right), two rows on each side in a square configuration (bottom left) and three rows on each side in a honeycomb configuration (bottom right). (b) Effects of changing the parameters of optical fiber on light intensity profile for fiber located at (x,y) = (0,0). Alterations described in (ii-v) are additive. (i) Intensity profile of a single fiber with diameter D = 0.5 mm, numerical aperture (NA) = 0.1, bending angle θ = 0º (perpendicular) and distance from the surface H = 5 mm. (ii) θ = 60º, (iii) D = 2.5 mm, (iv) NA = 0.4 and (v) H = 10 mm. (c) Light intensity profiles for different configurations, (i) one row at one side, (ii) one row at each side, (iii) five rows at each side in a square configuration and (iv) in a honeycomb configuration. Green boxes show the fontanelle area, and the middle green line is the US transducer axis. (d) (i) Evaluation metric, M (see Supplementary Section S1), for all ~124,000 possible probe configurations considered in this study, (ii) optimized configuration of fiber bundle on the probe side.

Fig. 4

Optical energy distribution through the fontanelle and evaluation of safety from thermal damage. (a) Fluence inside the tissue and significance of fontanelle, (i) the geometry used for Monte Carlo light simulations with three layers of the scalp (2 mm), skull (4 mm), and brain (20 mm), (ii) the same geometry with a 20 × 20 mm area of scalp to act as fontanelle, (iii) fluence inside the tissue for illumination through the skull and (iv) for trans-fontanelle illumination. (b) Normalized fluence decay in log scale (vertical dashed line corresponds to the bottom edge of the skull/ fontanelle as shown in (a-i and ii)). (c) Thermal studies: (i) cumulative equivalent minutes at 43 ºC (CEM43ºC); CEM43ºC model gives the equivalent exposure time with the same thermal effect as if the temperature was constant at 43ºC, (ii) temperature profile for different configurations of pulse repetition rates and energy levels after illuminating the tissue for 30 s; below the dashed line is considered as a safe choice.

Fig. 5

TFPAI probe resolution as a function depth. (a) Schematic of the experimental setup, including a hair phantom photograph captured by a 4 × objective on a light microscope (SME-F8BH, Amscope, CA, USA). (b) PA image of a hair. The dotted line shows the cross section from which the full-width half maximum (FWHM) is calculated. (c) 1D intensity profile obtained from the specified cross section. (d) 2D Gaussian fit of the intensity profile. (e) Axial and (f) lateral resolutions versus depth. OPO: optical parametric oscillator.

Fig. 6

PA versus US amplitude in measuring the severity of hemorrhages. (a) Schematic of the experimental setup. (b) Photograph of heparinized sheep blood sample diluted with saline suspended in gelatin phantom mimicking blood in CSF or early-stage intraventricular hemorrhage, and (c) photograph of heparinized sheep blood samples mixed with blended brain tissue in gelatin phantom mimicking intracerebral hemorrhage. (d) PA and US signal amplitudes from heparinized sheep blood diluted with saline in gelatin phantom. (e) PA and US signal amplitudes from heparinized sheep blood mixed with blended brain tissue in gelatin phantom. (f and g) Data points in (d) and (e) at lower concentrations are magnified. The inset in (g) is a magnified version of the specified box in (g). OPO: optical parametric oscillator, DAQ: data acquisition, CSF: cerebrospinal fluid.

Fig. 7

A feasibility study of TFPAI for detection of intracerebral hemorrhage in anex-vivomodel in sheep. (a) (i)The preparation of the ex-vivo sheep head included drilling of a 6 cm diameter cranial window in frontal bone and injection of blood-brain tissue mixture, (ii) experimental setup including sheep head, head holder, TFPAI probe, and optical fiber bundle, (iii) sheep head showing probe placement on induced cranial window, (iv) detail of probe with optical fiber bundle and placement at the surgically-created cranial window. (b) Hemorrhage detection image processing steps applied on PA images taken from the sheep brain ex-vivo, (i) PA image of the sheep brain before blood injection (40 µL), (ii) PA image of the sheep brain after blood injection, (iii) subtraction of the image in (i) from the image in (ii), (iv) binarize the resultant image, (v) overlaid hemorrhagic identified region on the US image. (c) The overlaid hemorrhagic area on US images. With increasing volumes of injected blood, the hemorrhagic area has increased. Volumes of blood injected were: (i) 20 µL, (ii) 40 µL, (iii) 60 µL, (iv) 80 µL, (v) 100 µL at 2.5 cm deep inside the brain tissue. (d) PA images of injected hemorrhage (20 µL) at different depths: (i) 0.5 cm, (ii) 1.0 cm, (iii) 1.5 cm, (iv) 2.0 cm, and (v) 2.5 cm overlaid on US images. (e) Actual pixel area of hemorrhages determined from PA and US images as a function of hemorrhage volume (20, 40, 60, 80, and 100 µL), images acquired at ~2.5 cm depth. (f) Measured fluence at different depths of brain tissue (0.5–3.0 cm) as a function of wavelength, ranging from 700 nm to 900 nm. (g) Normalized, averaged intensity extracted from US image (blue) and normalized, averaged, fluence-compensated intensity extracted from PA image (red), and corresponding fitted curves demonstrating quantification of the same hemorrhage (20 µL) at different depths. The fluence compensation curve is shown as an inset. SA: subarachnoid, CV: cortical vasculature, RV: right ventricle, LV: left ventricle, SB: skull base.

Fig. 8

Oxygen saturation measurement in-vitro. (a) Schematic of the oxygen saturation measurement and validation setup. (b) Experimental setup of in-vitro probe characterization. (i) Laser head, (ii) optical parametric oscillator, (iii) fiber bundle, (iv) energy meter (Gentec-EO, Canada), (v) automatic syringe pump, (vi) intralipid solution, (vii) TFPAI probe, (viii) capillary representing vessel, (ix) translation stage, (x) laptop for energy measurement and wavelength switching. (c) Homogenized sheep brain tissue was tested for use as a tissue-mimicking environment in in-vitro oxygenation study. Brain tissue mixture (i) whole phantom (initial phase) before removing air bubbles in solution generated from the blending, (ii) whole phantom (after removing bubbles), PA images (iii) before and (iv) after removing air bubbles. (d) Oxygen saturation of continuously flowing heparinized sheep blood within the range of 62–95% measured by TFPAI at depths of 0.5–5.0 cm with 0.5 cm increments; compared with and validated by the gold standard BGA measurement results. (e) Average error percentage. OPO: optical parametric oscillators, BGA: blood gas analyzer, DAQ: data acquisition.

Fig. 9

Validation of TFPAI vessel rupture localization and vasogenic edema detection using MB. (a) Schematic of the experimental setup for imaging the vessel rupture phantom and edema model phantom. (b) Steps for creating (i) vessel rupture phantom and (ii) edema model phantom (see Section 2.3.4 for a full description). (d) (i) injection of blood in the main vessel and imaged at 758 nm, (ii) reduction in PA image intensity when imaged at 690 nm, (iii) intensity increase in both main vessel and the rupture location due to presence of MB, imaged at 690 nm. The rupture location is annotated. (e) (i) Molar extinction coefficients from 250 to 800 nm for MB, Hb, and HbO₂. Dashed lines represent wavelengths used in these experiments, (ii) normalized MB concentrations over time for edema-like inclusion (with albumin) (red) and hemorrhage-like inclusion (no albumin) (blue) calculated through unmixing of Hb and HbO2 from MB .