Search Results

The MDHearing Smart Hearing Aids are self-fitting air-conduction hearing aids, intended to amplify sound for individuals 18 years of age or older with perceived mild to moderate hearing impairment. They are to meet the user's hearing needs. No Pre-programming or hearing test is necessary. The device is intended for direct-to-consumer sale and use without the assistance of a hearing care professional.

The MDHearing Smart Hearing Aid is a self-fitting wireless air conduction hearing aid system consisting of the Intricon Lumen 200B hardware and the MDHearing mobile application, which is an "app" compatible only with MDHearing devices of a Smart Hearing Aid product line, designed to interface with a user's compatible smartphone or tablet to personalize and manipulate the device and its settings. The wireless hearing aid incorporates microphones and a receiver encased in the behind-the-ear (BTE) hearing aid body, delivering amplified sound to the ear via standard thin tubing coupled to an earpiece for audio input into the ear. The hearing aid can be controlled wirelessly via Bluetooth Low Energy® using the MDHearing app or manually with on-device push buttons for changing volume and programs. The controls accessible through the MDHearing app and on the hearing aids are used to configure parameters, settings, and listening modes of the devices. The MDHearing Smart Hearing Aid is powered by a standard disposable size 312 zinc-air hearing aid battery.

The provided text describes the MDHearing Smart Hearing Aid and the studies conducted to demonstrate its substantial equivalence to a predicate device (Bose® Hearing Aid). The primary focus of the clinical performance testing was to show that self-fitting the device produced outcomes non-inferior to a professional fitting.

Here's a breakdown of the requested information based on the provided text:

Acceptance Criteria and Device Performance

The clinical performance validation aimed to demonstrate non-inferiority of the self-fit MDHearing Smart Hearing Aid compared to the same device professionally fit. The primary test metrics were user-reported (subjective) aided benefit using two standard questionnaires: the Abbreviated Profile of Hearing Aid Benefit (APHAB) and the 12-item short form of the Speech, Spatial and Qualities of Hearing scale (SSQ12). A secondary metric was speech-in-noise recognition using the QuickSIN test.

The text states: "The study results demonstrated that the MDHearing Smart Hearing Aid was non-inferior to fitting by a hearing care professional for both subjective and objective measures of hearing aid benefit."

The specific acceptance criteria (e.g., non-inferiority margins) are not explicitly stated as numerical thresholds in this document, but the graphical representations (Figure 2b and Figure 3b) show confidence intervals relative to a non-inferiority margin (-δ). The conclusion is that these intervals did not extend beyond the non-inferiority boundary.

Table of Acceptance Criteria and Reported Device Performance (Summary based on text):

Study Details:

1. Sample sizes used for the test set and the data provenance:

   • Clinical Performance Study (RCT):
     • Total participants: 64 adults (18 years or older with sensorineural hearing loss from mild to moderate degree).
     • Randomized 1:1 into two groups: 32 in "Professional-Fit" group and 32 in "Self-Fit" group.
     • Data provenance: Single-site clinical validation study. The country is not explicitly stated, but it's an FDA submission, implying a US study. The study was prospective ("randomized clinical validation study").
   • Reliability and Validity of Self-Fitting (in-lab):
     • 44 subjects (N = 88 ears) for the reliability measures (MAD in gain).
     • 58 ears (subset of professional-fit group) for comparison of user-preferred vs. clinician fine-tuned gain.
     • N=56 ears (28 participants) for REAR matching to NAL-NL2 targets.
   • Human Factors Study:
     • 20 untrained participants representing the intended user population.
   • Electroacoustic Performance (Non-Clinical):
     • Testing performed on the device itself, not on human subjects.

2. Number of experts used to establish the ground truth for the test set and the qualifications of those experts:

   • For the "Professional-Fit" arm of the clinical study, a clinician (presumably an audiologist or hearing care professional) programmed and fine-tuned the hearing aids "according to audiologic standards of care." The exact number and specific qualifications (e.g., years of experience) of these clinicians are not provided, only that they were "hearing care professionals" and "clinicians." Given it's a "single site" study, it likely involved a limited number of professionals.
   • For the self-fit group, the "ground truth" for comparison was the professional-fit group's performance. The subjective and objective outcome measures themselves served as a form of "ground truth" to compare the two fitting methods.

3. Adjudication method (e.g., 2+1, 3+1, none) for the test set:

   • Not applicable in the context of this hearing aid study. This type of adjudication is typically used for image-based diagnostic studies where human readers evaluate image findings and reach a consensus. Here, the outcome measures are quantitative (questionnaire scores, QuickSIN, real-ear measures) or direct observation of user interaction (human factors). The "professional-fit" outcome served as a benchmark, not an adjudicated ground truth in the sense of expert consensus on an interpretation.

4. If a multi reader multi case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance:

   • An MRMC study was not conducted. This was a randomized controlled trial comparing two fitting methods (self-fit vs. professional-fit) for a hearing aid, not an AI-assisted diagnostic task involving human readers. The "AI" component is the self-fitting algorithm within the device and app. The language used (effect size of human readers improving with AI) is typically for AI in diagnostic imaging.

5. If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:

   • The "self-fitting" mechanism itself is the core of the algorithm's standalone performance in providing an initial fit and allowing user adjustments. The clinical study implicitly evaluates this by comparing outcomes of the "self-fit" group (who primarily relied on the algorithm and their own adjustments) to the "professional-fit" group.
   • The "Reliability and Validity of the MDHearing Self-Fitting Method" section described in-lab testing where probe-microphone measures were taken after user adjustments, evaluating the consistency and approximation of the self-selected gain to professionally selected targets (NAL-NL2), this could be considered a form of evaluating the algorithm's output through user interaction.

6. The type of ground truth used (expert consensus, pathology, outcomes data, etc.):

   • The primary ground truth for the clinical comparison was the outcomes data from the "professional-fit" group, which served as the benchmark for demonstrating non-inferiority.
     • Subjective Outcomes: Patient-reported measures (APHAB, SSQ12).
     • Objective Outcomes: Speech-in-noise recognition (QuickSIN) and Real Ear Measures (REMs) against prescribed NAL-NL2 targets (an audiologic standard).
   • For the human factors study, the ground truth was the observed ability of participants to safely and correctly operate the device and app, guided by pre-defined task completion criteria.

7. The sample size for the training set:

   • The document does not explicitly state a separate training set size for the self-fitting algorithm. The NAL-NL2 fitting formula, which the device utilizes, is a well-established, validated prescriptive algorithm for hearing aid fitting, developed based on extensive audiological research and data. It's not a machine learning model that's "trained" in the typical sense on a specific dataset for this device. The app personalizes settings based on user input, which then applies this pre-established NAL-NL2 formula for the initial fit.

8. How the ground truth for the training set was established:

   • As noted above, the device uses the NAL-NL2 fitting formula, which is a widely accepted audiological standard, not a data-trained algorithm in the modern AI sense. Therefore, there wasn't a "training set" for this specific device's algorithm that required establishing ground truth by human experts. The "ground truth" for NAL-NL2 itself comes from decades of audiological research and clinical consensus on optimal gain targets for speech intelligibility. The device "applies personalized gain settings based on user input and fine-tuning. Utilizes NAL-NL2 fitting algorithm."

Ask a specific question about this device