Technological Advances in Universal Neonatal Hearing Screening (UNHS)

Abstract This paper presents a review of the major technological contributions in the field of universal neonatal hearing screening, during the last decade. Dedicated sections include references to the following : 1) the role of automated auditory brainstem responses (AABR) in screening; 2) the potential role of Auditory Steady State responses (ASSR) as a substitute of AABR; 3) the assessment of hearing threshold via recordings of distortion product otoacoustic emissions and the limitations of the clinical protocols to a hearing level < than 35dB HL; 4) the integration of multiple assessment protocols in one portable device, to be used in the intervention part of the hearing screening programme.


Introduction
In 2013, Otoacoustic Emissions (OAEs) celebrate a lifespan of 35 years (after the fi rst OAE publication by Kemp in 1978 (1). The most signifi cant contribution of OAEs is in the form of testing protocols in Universal Neonatal Hearing Screening (UNHS) programmes. Within the last decade, numerous new objectives have been presented in the UNHS fi eld such as: 1) the validity of automated screeners; 2) the quality of the automated OAE responses; 3) the estimation of hearing threshold; 4) the integration of numerous measurements in a portable automated device. To respond to the clinical demands of these objectives, several new methodologies have been introduced in clinical practice and the aim of this paper is to provide information on these technological trends.
While the main objective of neonatal hearing screening (NHS) is the identifi cation of infants with a hearing defi cit ( Ն 30dB HL), the objectives of a UNHS programme include the identifi cation of cases with hearing defi cits and, most importantly, intervention in terms of hearing improvement (hearing aids, cochlear implants) and neural rehabilitation. Therefore, the technologies developed for the intervention part of the UNHS programme exceed the limits of a simple identifi cation and offer estimates of the hearing status of the infant identifi ed with a hearing defi cit.
The sections below refer to the most important technological contributions in the area of UNHS, during the last 10 years.

Automated Auditory Brainstem Responses
In early 2002, the fi rst 4th generation OAE devices appeared in the market and offered the possibility to integrate information from different testing protocols such as automated OAE (AOAE) and automated ABR (AABR) responses. The combined screening protocols (AOAE ϩ AABR) targeted the identifi cation of auditory neuropathy cases, most prevalent in the neonatal intensive care (NICU) environment. Nevertheless, the presence of portable AABR equipment provided the possibility to conduct studies in real screening environments (and not in various Hearing, Balance and Communication, 2013;11: 104-109 Technological advances in UNHS 105 simulations in ideal ambient conditions) where the hearing threshold was assessed with both portable and clinical equipment. After the introduction of AABR in NHS, several issues became evident and among those were questions about the required screening times and screening costs. The latter is outside the objectives of this paper and will not be addressed. A study in the context of the regional NHS project CHEAP in Emilia-Romagna, Italy (2), provided evidence suggesting that in terms of timerequirements, the portable ABR and OAE technologies were converging to the same values. The data from the above study suggested: 1) the average time for AOAE responses is less than 10 s in a cooperative subject and less that 120 s (2 min) in non-cooperative subjects; 2) test times of AABR in cooperative subjects were less than 120 s, while uncooperative subjects were tested within 10 min (per ear). While it takes some minimum expertise to properly handle and position the OAE probe, the ABR electrode placement presents more complications especially in cases where the subject shows high electrode impedance. In the latter case AABR testing is diffi cult to complete and the test times are unavoidably longer.
Theoretically this two-stage approach (i.e. AOAE ϩ AABR) eliminates the risk of not identifying infants with auditory neuropathy and ensures that the screening sensitivity is high. In contrast to these hypotheses, data from an American study (3) suggest that this is not the case. The study assessed information from 86,634 infants and for the infants who were screened for hearing loss using a typical twostage OAE/A-ABR protocol, approximately 23% of those with permanent hearing loss at 8 -12 months of age would have passed the AABR. The data suggest that stringent criteria should be incorporated in the fi nal evaluation of the current OAE and ABR automated devices.

Auditory Steady State Responses in neonatal screening
Both OAE and ABR technologies utilize electrical clicks as stimuli and the acquired information is clearly more related to the audiometric frequencies of 1 and 2 kHz. Within this context, there has been speculation as to whether or not other measurement technologies could be used in a fast hearing assessment of neonates, children and adults. A group of similar electrophysiological measurements to OAEs and AABR includes electro-cochleography (EcoG), Middle Latency (ML) and Steady State Responses (SSR). From this group the latter category has shown interesting characteristics due to the fact that by alternating the modulation frequency (i.e. increasing it) of the stimuli one can obtain responses from the auditory cortex (low modulation frequencies around 40 Hz) or from the brainstem (4 -6). The SSR protocol has already passed to an automated one (ASSR) and for the last 10 years numerous publications have been devoted to threshold estimation via the ASSR technique. The ASSR protocols have been greatly optimized (7) and the SSR responses are detected in the frequency domain by robust probabilistic algorithms.
In 2002, Conne-Wesson et al. (4) proposed the use of ASSR as a hearing screening tool, with the objective that ASSR could substitute for the AABR. A few reports have since been available (8,9), indicating a good agreement between ASSR and AABR at 2 kHz and various differences at 0.5, 1 and 4 kHz. Most studies recommended the use of the SSR technique in the clinic but the point of substituting the AABR with ASSR is not fully supported by the available data.
The factors that affect the AABR (ambient noise and electrode impedance) interfere with the ASSR recordings as well. In order to resolve these issues Vivosonic presented in 2008 a new line of devices using pre-amplifi ers at the level of the scalp electrodes (called amplitrodes), which suppress the level of ambient noise and provide very clean AABR and ASSR traces. It remains to be seen how these electrodes will be intergrated in the normal clinical reality since the pre-amplifi ers require electrical energy, which translates into changing batteries after a limited number of tests.
In the context of neonatal screening, an ASSR screening protocol might target initially a few frequency points (i.e. 1 and 2 kHz or 2 and 4 kHz) that show immunity to ambient noise (Figures 1 and 2). Figure 1. ASSR response from a well baby who was crying using the Audera device from Viasys. The lowest tested frequency was not available due to noise. The length of the testing procedure was 22 min (14 min longer than the successfully completed AABR test). Despite the theoretical noise immunity at 2.0 and 4.0 kHz the size of the error bars indicates that the measurements are too variable to be considered. The ' ϫ ' symbols indicate the mean threshold level of the measurements. Figure 2. ASSR response from a well baby using the Audera device. The length of the test was also longer than the AABR (16 vs. 7 min). The AABR proposed a REFER probably due to conductive complications suggested by the AASR outcome. In this case the 2.0 and 4.0 kHz frequencies show good noise immunity (suggested by the small size of the error bars). The ' ϫ ' symbols indicate the mean threshold level of the measurements.
Nevertheless, the ASSR protocol requires signifi cant optimizations before becoming a member of the neonatal hearing screening battery of tests. Recently, a study (10) presented data on the relationship between ABR, ASSR estimates and data from Conditioned Orientation Responses (COR), a technique widely used in the intervention phase of many UNHS programmes. The data suggested a very good relationship between the outcomes of the ASSR and COR techniques, with the ASSR data being closer to the ABR estimates. Data from large-scale studies in this direction (i.e. comparing ASSR with other protocols) could support this hypothesis and eliminate the use of ABR and COR in the intervention phase of a UNHS programme.

Threshold estimation via DPOAE measurements
An interesting challenge for otoacoustic emissions has been the relationship between the amplitude of the OAE response and the hearing threshold (11 -13). For cases where no conductive losses are present there is good agreement between OAEs and the hearing threshold. In such cases input-output distortion product OAE (DPOAE) protocols may offer more information (11,14 -16). Besides the relationship to pure tone thresholds, DPOAE I/O functions provide an estimate of the compression related to the outer hair cell amplifi er. Data supporting this hypothesis are available from animal studies where the hearing of the animals was impaired with acute furosemide intoxication (17), and human studies with subjects suffering from cochlear hearing loss (14,18,19). In these studies the slope of the DPOAE I/O function increased with increasing hearing loss, revealing a loss of compression of the outer hair cell amplifi ers. In this context, by using numerous combinations of I/O DPOAE recordings, one can obtain very precise information relating to the status of the cochlear amplifi er (16,20). Extrapolated DPOAE I/O functions were constructed from neonates to estimate pure tone threshold levels and the corresponding cochlear compression values (21). The estimated hearing threshold was found to be increasing within the early postnatal period (average age, three days), predominantly at the higher frequencies, and to be normalized in a follow-up measurement (after four weeks). However, the slope of the DPOAE I/O functions obtained in the fi rst and second measurement was unchanged, revealing normal cochlear compression. Consequently, these fi ndings were interpreted as temporary conductive hearing losses due to the presence of amniotic fl uid and/or Eustachian tube dysfunction. In this clinical scenario, especially during the fi rst days of life, a hearing screening test may lead to false positive results due to a temporary conductive hearing loss. The use of the slope of DPOAE I/O functions could be used as an index of conductive losses, which might result in fewer false positives and in less time spent on audiological clinical diagnostics. According to the data of Jenssen et al. (21), the values of the DPOAE slope can discriminate and differentiate conductive from sensorineural hearing losses. In addition, DPOAE I/O functions have been reported to be correlated with loudness (20), so DPOAE I/O information would also offer the potential of providing information for basic hearing aid fi tting.
Research fi ndings from Janssen et al. (21), and Gorga et al. (20), have been commercialized in a device called Cochlea-Scan (22) by Natus. Hearing threshold can be extrapolated up to values relative to 50dB HL in the frequency range from 1.5 to 6 kHz. Figure 3 shows a typical hearing threshold profi le and the corresponding Cochlea-Scan mediated estimation of hearing threshold. At present, the Cochlea-Scan device offers a platform for a third generation OAE testing (TEOAEs, DPOAEs), I/O DPOAE estimation with hearing threshold extrapolation.
Further analyses (23,24) on the effi cacy of the Cochlea-Scan DPOAE algorithm, relating hearing threshold data and Cochlea-Scan estimated thresholds from a group of adult sensorineural cases, suggested a different scenario than the one proposed initially by Janssen (21). In the Hatzopoulos et al. (24) study, behavioural and Cochlea-Scan data were analysed with logistic regression models in order to The seconds depicts the DPOAE audiogram, the x-axis shows frequency in kHz and the y-axis shows hearing level in dB HL. The third panel shows the DP-gram for each frequency, and from these it is possible to estimate the overall DPOAE slope. The acronym NA means no threshold estimation was possible at the tested frequency.
fi nd the probability ( Յ 0.9) of a robust DPOAE response at 2, 3, and 4 KHz. The data suggested that the maximum behavioural levels where valid DPOAEs could be detected were equal to 32.8, 21, and 34dB, respectively. For normal hearing adults the detection levels were lower. Figure 4 depicts the data for 2 kHz, suggesting a detection threshold approximately at 15dB HL. While it is still possible to have a detection threshold as high as 50dB HL, the corresponding probability falls below 30% and as such limits the usefulness of the Cochlea-Scan protocols in a UNHS programme.

Integration of multiple hearing assessment protocols into an automated device
The success of the NHS screening practices challenged another area of paediatric audiology -the area of school screening. Data from large-scale screening programmes, as in Poland, suggested that in this area different protocols could be applied than in UNHS programmes, with the emphasis on pure tone behavioural responses, tympanometry and ABR (25 -27). The OAEs were found the least effective tool in the battery of screening tests, suffering mainly from the ambient noise present in schools.
A number of OAE manufacturers (Natus, PathMed) proposed hand-held devices (fi fth generation OAE equipment) capable of testing subjects with behavioural protocols, OAEs/AOAEs, AABR and ASSR. A tympanometry assessment has not appeared to date due to complications in the probe of the device. Recently PathMed proposed a new device (Sentiero) capable not only of the above protocols but also of protocols for speech audiometry. Such a device could be easily implemented in the intervention part of the UNHS programme and eventually in the area of neural and auditory rehabilitation.

Conclusions
In the last 10 years signifi cant advances have been made in the integration of various protocols and various approaches in the UNHS strategies. The most important contribution is in the area of Auditory Steady State Responses that have been shown to be well correlated with other measures in  audiology such as the AABR, ABR, OAEs and COR. The current technological trends call for an integration of even more protocols and algorithms in a hand-held device. The clinical robustness and response quality of these new entries are yet to be evaluated.

Declaration of interest:
The authors report no confl icts of interest. The authors alone are responsible for the content and writing of the paper.