Stability and Test-Retest Reliability of the Individual Acoustic Reference for Within-Subject Analysis of Hearing Device-Processed Speech: The SpettroREM Method.

@article{burdo2026stability,
  title = {{Stability and Test-Retest Reliability of the Individual Acoustic Reference for Within-Subject Analysis of Hearing Device-Processed Speech: The SpettroREM Method.}},
  author = {Sandro Burdo and Edoardo Carini and Giuseppe Impalà and Davide Brotto},
  journal = {Audiology research},
  year = {2026},
  month = oct,
  doi = {10.3390/audiolres16050158},
  url = {https://europepmc.org/article/MED/42930356},
  note = {CC BY},
}

TY  - JOUR
TI  - Stability and Test-Retest Reliability of the Individual Acoustic Reference for Within-Subject Analysis of Hearing Device-Processed Speech: The SpettroREM Method.
AU  - Sandro Burdo
AU  - Edoardo Carini
AU  - Giuseppe Impalà
AU  - Davide Brotto
PY  - 2026
DA  - 2026/10/11
JO  - Audiology research
DO  - 10.3390/audiolres16050158
UR  - https://europepmc.org/article/MED/42930356
AB  - Objective: Modern hearing devices modify speech beyond simple amplification, whereas Real-Ear Measurement (REM) primarily quantifies the sound level delivered to the ear as a function of frequency. SpettroREM was developed to characterize spectral and temporal signal structure by comparing each aided recording with the corresponding unaided recording, termed the Individual Acoustic Reference (IAR). This study evaluates the stability and test-retest reliability of the unaided reference and characterizes the external recording pathway; it does not assess aided-unaided differences. Design: The descriptor framework was developed from preliminary auditory and visual analysis of four unaided ear-canal recordings. Established acoustic measures available in Praat were progressively extended to predefined speech segments, local measurements, derived spectral and temporal measures, vowel geometry, envelope dynamics, and speech contrast. The resulting exploratory library comprised 44 measures; 37 were designated as the principal descriptor set and organized into six complementary domains, whereas seven derived or partially overlapping measures were retained as exploratory outputs. REM recordings acquired using AURICAL FreeFit (Natus) were processed through a semi-automated Praat workflow. Descriptor variability was characterized using descriptive distributional measures and, where appropriate, coefficients of variation. Test-retest reliability was assessed using ICC(A,1) with 95% participant-level bootstrap confidence intervals, together with mean absolute differences, Bland-Altman limits of agreement, and repeatability coefficients. The external recording pathway was assessed through electrical loopback and acoustic comparison tests. Study Sample: This study included two complementary datasets: 26 unaided ear recordings for baseline descriptor characterization and bilateral test-retest recordings from 25 participants. The primary reliability analysis included 24 right ears because one right ear was excluded owing to tympanic membrane perforation. Test-retest recordings from all 25 contralateral left ears were analyzed separately as a within-sample sensitivity analysis to assess the robustness of the principal reliability pattern to the choice of ear used for the primary analysis. Right- and left-ear observations were not pooled. Results: Global spectral descriptors showed limited variability and the most consistent test-retest agreement in the primary right-ear analysis. In the exploratory post hoc tier classification, the seven Core descriptors showed ICC(A,1) values ranging from 0.906 to 0.984, with the lower bound of the 95% confidence interval remaining ≥0.75 for all Core descriptors. Mean absolute test-retest differences for the LTASS bands were 0.31 dB at 0-1 kHz, 0.50 dB at 1-4 kHz, and 0.88 dB at 4-8 kHz. The contralateral sensitivity analysis reproduced high repeatability of the three LTASS band levels, with ICCs ranging from 0.835 to 0.940, although it did not constitute an independent replication. Repeatability was less consistent for several recording-level-dependent, segmental, vowel-derived, and temporal-envelope descriptors, some of which were occasionally non-evaluable because of insufficient signal energy. Electrical testing showed no relevant spectral alteration or clipping in the cable-audio interface-Praat segment under the tested settings. External-recording repeatability was high across the primary mono-cable validation and the secondary stereo bridge experiment (r = 0.976-0.998), but frequency-dependent differences were observed between the OTOsuite display and the simultaneous external recording. The two acquisition outputs should therefore not be considered numerically interchangeable. Conclusions: Under standardized acquisition conditions, selected properties of the unaided IAR, particularly its global spectral organization, can be measured with good test-retest repeatability. Technical assessment supports internally consistent measurements using the tested AURICAL FreeFit-mono TS cable-external audio interface-Praat pathway but does not establish absolute spectral accuracy or numerical equivalence across acquisition outputs or platforms. The Core, Candidate, and Secondary classification is exploratory, post hoc, and requires confirmation in an independent sample. This study provides a preliminary reliability assessment of the unaided reference and its development-derived 37-descriptor library; it does not validate the complete SpettroREM method. Direct aided-unaided studies are required to establish sensitivity to device-induced transformations, aided-recording repeatability, clinical and perceptual relevance, and the most informative non-redundant descriptor subset.
SN  - 2039-4330
SN  - 2039-4349
ER  -