Neonatal phototherapy radiometers:

Current performance characteristics and

future requirements
Author links open overlay panelDouglas McG.Clarkson RuthNicol PhillipChapman
a b c

Show more

https://doi.org/10.1016/j.medengphy.2013.11.001Get rights and content

Abstract
Hand held radiometers provide a convenient means of monitoring the
output of neonatal phototherapy treatment devices as part of planned
programs of device maintenance and output monitoring. It was
considered appropriate to determine the wavelength and angular
response of a selection of such meters and compare their indicated values
with that derived from spectral analysis of phototherapy light sources.
This was undertaken using a Bentham DMc150 double grating
spectroradiometer and a series of 10 nm band pass optical filters in the
range 400–640 nm used in conjunction with a fiber optic light source.
Specific meters investigated included a GE Biliblanket Light Meter II, a
NeoBLUE radiometer and a Bio-TEK radiometer 74345 device.
Comparisons were made of measurements made using the hand held
meters and the Bentham DMc 150 system for a range of neonatal
phototherapy treatment devices. The use of such meters is discussed in
relation to applicable equipment standards and recommendations of
intensive phototherapy from clinical groups such as the American
Academy of Pediatrics and a specification for a spectroradiometer based
measurement system is proposed.

Keywords
Neonatal phototherapy
Hyperbilirubinemia

Neonatal jaundice

1. Introduction

1.1. Phototherapy radiometers

It was considered appropriate to compare output measurements of
neonatal phototherapy equipment made using a selection of hand held
meters with determinations of device irradiance using a Bentham DMc
150 spectroradiometer (Bentham Instruments, Reading, UK). Table 1
indicates the specific phototherapy meters investigated.
Table 1. Details of neonatal phototherapy meters investigated.
The bandwidth of such meters is typically identified as the wavelength interval
corresponding to 50% of meter wavelength sensitivity which is principally
derived from the combined response of the filter element and the specific photo
detector device utilized. Units of measurement used with phototherapy meters
investigated include μW cm  nm and μW cm , where a value of μW cm can be
−2 −1 −2 −2

converted to μW cm  nm by dividing the meter reading in μW cm by the

−2 −1 −2

bandwidth of the meter. Similarly a value in μW cm  nm can be converted to

−2 −1

μW cm by multiplying by the bandwidth in nm of the meter. In the example of

−2

the Bio-TEK Model 74345, a value of 480 μW cm would be equivalent to−2

10.9 μW cm  nm for a bandwidth value of 44 nm. The standards document BS

−2 −1

EN 60601-2-50:2009 [1] relates to irradiance values within the wavelength range

400–550 nm as having the potential for neonatal phototherapy.

1.2. Clinical framework

Previous guidelines of the American Academy of Pediatrics [2] referenced an
average level of 30 μW cm  nm within the wavelength band 430–490 nm –
−2 −1

equivalent to 1.8 mW cm within the active bandwidth as part of ‘intensive’

−2

phototherapy. A recent revision of this policy [3] references an average level of

30 μW cm  nm within the wavelength band 460–490 nm – equivalent to
−2 −1

0.9 mW cm within the active bandwidth. Maisels and McDonagh [4] identify
−2

specific pathways of bilirubin metabolism and indicate that the most effective
wavelengths are probably between 460 nm and 490 nm through no specific
evidence is cited for this observation. McDonagh and Lightner [5] describe the
complex modes of photo disruption of bilirubin in vivo, indicating the numerous
factors which distinguish in vitro from in vivo wavelength responses. A recent
study [6] has indicated a linear response of the rate of serum bilirubin reduction
with delivered light dose, with no indication of a ‘plateau effect’ with
progressively higher levels of incident light levels.

1.3. Models of radiometer response

With increasing clinical preference for ‘greener’ wavelengths for neonatal
phototherapy, it is instructive to derive values of wavelength functions of
phototherapy devices within the wavelength bands 430–490 nm and 460–490 nm
which are independent of specific commercially available phototherapy
radiometers.

A Gaussian response function G (λ) with peak sensitivity at 460 nm and 50%
430–490

bandwidth at 430 nm and 490 nm can be empirically described as

Similarly a Gaussian function G (λ) with peak sensitivity at 475 nm and
460–490

50% bandwidth at 460 nm and 490 nm can be empirically described as

In addition, ‘top hat’ functions TH (λ) and TH

430–490(λ) with unity value of
460–490

response within the specific inclusive wavebands were also identified. Fig.
1 indicates characteristics of TH (λ), G (λ) and G (λ).
460–490 430–490 460–490

Fig. 1. Specific functions of ‘top hat’ 460–490 nm, G 430–490 (λ) and G 460–490 (λ). The function
TH (λ) is not shown.
430–490

These functions were used to derive components of irradiance within the

specific wavelength functions for specific neonatal phototherapy devices
as indicated subsequently in Table 4. Functions such as TH (λ) and 430–490

TH (λ) can be directly measured by spectroradiometer devices such as

460–490

the Bentham DMc150 system, where a value of irradiance is determined

at individual wavelength values.

2. Method
2.1. Determination of angular response of phototherapy
radiometers
The angular response of the indicated meters and the Bentham DMc150
system with a D6 detection head were determined using light from a fiber
optic light source with a 4 mm cross section presented to the detector
surface at angles from −90° to +90° at intervals of 10°. The 4 mm fiber
optic light source provided a convenient means of generating high levels
of optical output within the wavelength sensitivity range of the
phototherapy meters. The distance between the tip of the fiber light
source and the surface of the specific detector was 22 cm. This distance
was identified as a compromise between allowing the phototherapy
meters to register a meaningful signal from the light source (shorter
distance better) and allowing improved angular resolution of
measurements (greater distance better).

The factor f [7], [8] is used as a ‘quality factor’ for determination of the
2

angular response of optical detectors. The directional error f for values of

2

α (angle of incidence to normal) in range 0–85° can be expressed as a

percentage:

where this value is normalized to the value 2 cos(α) sin(α) summed over the same
angular range.

2.2. Determination of wavelength response of phototherapy

radiometers
Fig. 2 indicates the filter housing used with the fiber optic light source where the
plane of measurement is either the contact surface of the meter being tested or the
open aperture of an integrating sphere (IS3-ODM, Bentham Instruments Ltd.,
Reading, UK) of aperture 12 mm. A range of nominal 10 nm bandwidth filters
(Knight Optical (UK) Ltd., Harrietsham, UK) of outer diameter 12.5 mm in
10 nm steps from 400 nm to 540 nm and then at 20 nm intervals between 540 nm
and 640 nm was used to determine the spectral response of the phototherapy
radiometers. The wavelength accuracy of the Bentham DMc150 was checked
with an HG-1 Mercury-Argon lamp (Ocean Optics Inc., Dunedin, USA) and the
uncertainty in wavelength value over the range 400–550 nm was estimated as
±1 nm. The spectral resolution of the Bentham DMc150 system – i.e. its intrinsic
ability to resolve spectral information, was estimated to be 1 nm based on the use
of an 0.33 mm entrance slit to the monochromator elements and an 0.33 mm exit
slit to the photomultiplier detection unit. Irradiance measurements made by the
Bentham DMc150 system were calibrated using a Bentham CL-6 tungsten lamp
with traceability to national standards.

Fig. 2. Measurement configuration of fiber optic light source, bandpass optical filter
and plane of measurement.

2.3. Measurements of neonatal phototherapy lamps

Specific measurements of light output of phototherapy units were undertaken
using the Bentham D6 detection head attached to a fiber optic cable and also
using the various phototherapy radiometers – recorded as measured values
Meter . Indicated distances of measurement for specific products were in
obs

accordance with the manufacturer's instructions for use.

Values of meter reading, Meter , of the various radiometers for specific
Est1

phototherapy light sources can be considered as the sum of contributions within

specific wavelength bands as indicated in Eq. (4) for center wavelengths of such
wavebands in the range 400–540 nm

where i(j) is the irradiance in a specific waveband as determined from the

Bentham spectral data and values of C(j) are coefficients of meter sensitivity
within the specific waveband. Eq. (4) assumes that the irradiance is uniform
across the entire sensitive area of each detection device.
For the set of optical filters, the values of relative spectral sensitivity R(λ)
for each waveband can be determined by assuming that the value of the
corresponding meter reading, Meter(j), is associated with the total
irradiance, i(j), transmitted by each filter as indicated in Eq. (5). The
geometry of the measurements with the filter in place as indicated in Fig.
2 does not, however, equate to the geometry of the exposed detector to a
uniform lamp source as described in Eq. (4)

At a given waveband centered at 430 nm, however, the meter reading can

be considered to include the sum of contributions from wavebands on
either side of the peak wavelength of 430 nm as indicated in Eq. (6),
where the main contribution is from the sensitivity at 430 nm but with
contributions from 410 nm, 420 nm, 440 nm and 450 nm. This is due to
the intrinsic spectral characteristics of the filters used

Values of terms (R /R ), (R /R ), etc., in the square brackets in Eq. (7)

410 430 420 430

were derived from values of R(j) using Eq. (5) and the corrected value of
R calculated. This process was undertaken for the range of coefficient
430

values. In this process the value of central irradiance is reduced but with
inclusion of contributions from other terms in adjacent wavebands. This
correction factor did not, however, significantly alter the wavelength
sensitivity characteristics of the meter devices. For each meter device, the
normalized wavelength response Resp_norm(j) with correction for
sidebands was determined for each waveband.
A estimation of meter reading, Meter , was derived as indicated in Eq.
Est2

(8) using the basic assumptions relating to the measurement function of

the radiometer devices
where Resp_norm(j) is the value of relative meter sensitivity in specific
wavebands (maximum value unity), i(j) is the measured spectral irradiance using
the Bentham DMc 150 system within specific wavebands and BW is the Meter

bandwidth of the radiometer. The bandwidths of the GE Biliblanket Light Meter

II and the NeoBLUE meter were assumed to be 60 nm and 62 nm, respectively.
For the Bio-TEK radiometer, the value of BW has effectively unity value since
meter

the meter units for this device are μW cm . This function is essentially deriving
−2

the effective irradiance within the active bandwidth of the phototherapy

radiometers based on the measured relative spectral response of the radiometer
meters and the spectral irradiance of the phototherapy lamp systems.

3. Results

3.1. Angular response

Fig. 3 indicates the angular response of the various detector devices relative to
the absolute cosine response. The cosine response of the Bio-TEK meter deviates
most from the theoretical cosine response due to the use of a recessed aperture
which restricts detection of light at increasing angles of incidence. Thus for the
Bio-TEK device at an angle of incidence of 60°, the detected signal would be
<50% of a ‘perfect’ cosine response and there is essentially no detected signal for
angles greater than 70°. The impact of this angular response on the actual meter
reading would depend on the relative angular distribution of light incident on the
detector.
Fig. 4 indicates the trend of magnitude of f % error based on CIE 69 [7]
2

with increasing angle to the vertical where, for example, the cumulative %
f value at 50% includes the sum of all contributions between 0° and 50°.
2

Cumulative values of f % error were calculated using Eq. (3) with

2

interpolation of points between 10° interval values. The value of the

NeoBLUE f % response tracks the f % response of the Bentham D6 until
2 2

around 40° after which its performance diverges. The Bio-TEK

consistently deviates with increasing angle to the normal. Table 2
summarizes values of f % for the specific measurement devices. While
2

values of f % are always positive, the interpretation of Fig. 3 is that all

2

deviations from the perfect cosine response of the Bio-TEK device

decrease the detected value. The measurement technique, however, is
limited by the dynamic range of measurements and the measurement
resolution of the NeoBLUE and the GE Biliblanket Light Meter II with the
available light source.
3.2. Meter wavelength response
Fig. 5 indicates the calculated relative spectral response, Resp_norm(j), of
the three phototherapy radiometers with correction for contributions
from filter ‘sidebands’ as indicated in Eq. (6). It was noted that the Bio-
TEK meter would saturate at filter outputs at 450 nm and 460 nm for
which the GE meter would indicate values of 5.4 μW cm  nm and
−2 −1

5.2 μW cm  nm and the NeoBLUE meter values of 5.2 μW cm  nm and
−2 −1 −2 −1

8.5 μW cm  nm . The meter devices registered essentially no sensitivity in

−2 −1

the range 540–640 nm.

3.3. Comparison of values of Meter and Meter obs Est2

Values of Meter are indicated in Table 3 and based on the

Est2

characterization of device sensitivity outlined in Eq. (8). Values of Meter Est2

for non-contact geometry were usually higher than values of Meter for obs

the GE Biliblanket Light Meter II and the NeoBLUE hand held radiometer
devices though typically within a margin of 10%. The lower values of
Meter relative to Meter , could have been introduced through the loss of
obs Est2

sensitivity due to the cosine response of the detectors.

Table 3. Comparison of output from a range of phototherapy devices as
measured using hand held meters and specific wavelength ranges using the
Bentham DMc150 radiometer (OR = over range where the detected signal
exceeds the meter's maximum displayed value and a reading of the
phototherapy device cannot be obtained).
The Bio-TEK meter did not indicate as good agreement between Meter and obs

Meter for non-contact lamp systems. Some observed values of Meter for the
Est2 obs

Bio-TEK unit were higher than Meter (Giraffe Photospot (40%), NeoBLUE
Est2

(low) 64%, GE Lullaby (low) 34%). The cosine response of the Bio-TEK meter
would tend to result in values of Meter less than that of Meter . The Bio-TEK
obs Est2

device was not able to record the irradiance value of several of the phototherapy
lamps since the indicated irradiance value was greater than its maximum
displayable value.
Measurements made on fiber optic pads showed more variation between
phototherapy meters and the Bentham DMc 150 measurements. A component of
this variability relates to the difficulty of selecting identical treatment areas on
the surface mats using different detection devices.
3.4. Mapping of phototherapy lamp outputs to defined bandwidth
functions
The predicted irradiance in specific waveband outputs as indicated in Table 4
was derived from the measured spectral output of the phototherapy devices and
the previous identified wavebands indicated in Fig. 1. Values in units of
μW cm  nm were in turn calculated using the bandwidth values of the respective
−2 −1

devices.
Table 4. Estimated irradiance within spectral functions TH430–490(λ), G430–490(λ), TH460–490(λ)
and G (λ) based on Bentham DMc 150 measurements with values indicated in
460–490

units of μW cm  nm and mW cm .

−2 −1 −2
Table 4 provides an indication of the spectral content for specific defined
wavelength functions in the context of the American Academy of Pediatrics
guidelines and which are essentially independent of specific proprietary
phototherapy radiometers. This indicates the advantage of describing delivered
phototherapy optical radiation within the context of well defined spectral
information.

4. Discussion
Hand held phototherapy radiometers provide a useful role in measuring outputs
in order to maintain treatment levels within product specifications. Their primary
role, however, is typically for use with specific phototherapy light sources. It is
customary in device validation procedures of phototherapy radiometers to present
a narrow bandwidth light source of known irradiance to the unit being tested at
normal incidence and at the peak of wavelength sensitivity. Such a process is
therefore neither checking the spectral response nor the cosine response function
of the device, both of which are elements which can significantly affect the
performance of the device.

The light sensitive area of each hand held meter detector was not directly
investigated. It is likely, however, that the relative sensitivity of such devices
varies within the light sensitive area of each device and which would affect the
measurement of light fields of varying uniformity.

Manufacturers of neonatal phototherapy devices should provide information of

the typical spectral output at standard delivery geometry as absolute irradiance
values at 1 nm intervals in units of mW cm within a spectral window of at least
−2

400–550 nm. This would allow spectral irradiance in units of mW cm within

−2

specific wavelength intervals of clinical significance to be more clearly

expressed. In addition, it is desirable that information on the relative irradiance
values is provided within a grid system, such as described by Vreman et al. [10].
While the meters tested had wavelength sensitivity within the wavelength range
460–490 nm as recently referenced by the American Academy of Pediatrics [3],
these meters have also significant sensitivity outside this wavelength range. In
addition, levels of clinical effectiveness have to be considered in terms of the
effective area of exposure, where the effective treatment area may also be limited
by the size of the blanket and the level of contact with the neonate surface. This
is an area where further developments in metrology are awaited.
Spectral outputs of phototherapy devices continue to be poorly understood. Table
4 indicates how specific spectral data output can be expressed in relation to the
defined wavelength functions independent of commercially available
phototherapy radiometers. Such estimations can only be made reliably using
spectroradiometer systems of appropriate specification and accuracy.
Based on the clinical importance of appropriate administration of neonatal
phototherapy light dose, it is identified that spectroradiometer systems of
appropriate design should be employed as a means of providing the highest level
of possible accuracy of delivered light energy. It is possible to identify some of
the desirable characteristics of a system for more accurate neonatal phototherapy
dosimetry as indicated in Table 5. Such a device would be used as the reference
standard for clinical investigation of effectiveness of neonatal phototherapy and
also with a role for routine verification of treatment regimes. A feature of such
instrumentation would be the ability to integrate dose levels in units of J cm −2

within identified wavebands. It is considered important to raise the standards of

clinical measurement of neonatal light therapy in an environment where much
higher significance is attached to the measurement of patient parameters – such
as in this context values of serum bilirubin. Devices such as the STS
Microspectrometer (Ocean Optics Inc., Dunedin, USA) with a spectral resolution
of 1.5 nm indicate the availability of potentially suitable spectral measurement
components. The key design element of such a proposed instrument would be the
light collection/delivery system to the micro spectrometer element to ensure a
satisfactory cosine response.
Table 5. Desirable characteristics of a system for neonatal phototherapy dosimetry
based on a spectroradiometer system.
It is acknowledged that the cost of such spectroradiometers would be
significantly greater than commonly used hand held radiometers. This
cost would, however, still represent a small component of the overall
equipment cost within a neonatal intensive care facility and could
potentially be significantly less than the equivalent cost of sub optimal
clinical care.

Oláh et al. [11] and Csoma et al. [12] highlight the potential long term
risks of neonatal blue light therapy with respect to development of
melanocytic naevus and also ophthalmic changes associated with
structures such as iris freckles, iris nevi and choroidal nevi. Previous
studies such as that of Mahé et al. [13] have identified no causal link
between prevalence of atypical nevi and previous exposure to neonatal
phototherapy. The more recent study indicated an increased occurrence
of these characteristics within twins who had received blue light
phototherapy at birth. While it is acknowledged that the risks of
hyperbilirubinemia are immediate and in extreme cases irreversible, the
authors indicate the need to provide treatment only where indicated. In
this context, however, this implies an increased importance for accurate
determination of delivered levels of phototherapy which includes spectral
information.