FDA 510(k) Search - By Innolitics (SaMD and AI Experts)

Sonu is indicated for the relief of moderate to severe nasal congestion due to allergic and non-allergic rhinitis. Sonu is a treatment to be used at home by individuals 22 and older.

Device Description

Sonu is a non-invasive, over-the-counter (OTC) device designed for the relief of moderate to severe nasal congestion due to allergic and non-allergic rhinitis at home. Sonu consists of an adjustable headband (Sonu Band) with integrated acoustic bone-conduction transducers, a USBC Charging Cable, and a smartphone application (Sonu iOS App) that connects to the Sonu Band.

AI/ML Overview

Here's an analysis of the Sonu device's acceptance criteria and the study proving it meets them, based on the provided text:

1. Table of Acceptance Criteria and Reported Device Performance

Acceptance Criteria Category	Specific Criteria	Reported Device Performance
Primary Safety	Absence of Serious Adverse Events (SAEs) related to the device or treatment.	Achieved: 0 Device or Procedure-Related SAEs reported in 52 subjects over 1234 total treatments.
Secondary Safety	Absence of Adverse Events (AEs) related to the device or treatment.	Achieved: 0 Device or Procedure-Related AEs reported in 52 subjects over 1234 total treatments.
Primary Effectiveness	Statistically significant improvement in nasal congestion sub-score of TNSS, comparing baseline congestion to daily congestion averaged over 2 weeks (Sonu vs. Sham), exceeding a sponsor-defined Minimal Clinically Important Difference (MCID).	Achieved: Mean change in nasal congestion sub-score was -0.87 for Sonu Therapy vs. -0.44 for Sham (p=0.008). This difference (0.43, adjusted to 0.47 with 95% CI 0.15, 0.79, p=0.005) was statistically significant and exceeded the MCID. The nasal congestion sub-score decreased by 33% in Week 1 and 44.5% in Week 2 for Sonu Therapy vs. 20.3% and 19.3% for Sham respectively, with statistical significance by Week 1 (p=0.030) and Week 2 (p=0.008).
Secondary Effectiveness	Statistically significant improvement in 24-hour reflective TNSS comparing baseline TNSS to daily TNSS averaged over 2 weeks (Sonu vs. Sham).	Achieved: Mean change in composite TNSS was -2.85 for Sonu Therapy vs. -1.32 for Sham (p=0.027). The difference (1.53, adjusted to 1.80 with 95% CI 0.46, 3.13, p=0.009) was statistically significant. TNSS decreased by 38% in Week 1 and 50.1% in Week 2 for Sonu Therapy vs. 20.2% and 22.5% for Sham, reaching statistical significance by Week 2 (p=0.02).
Biocompatibility	Patient-contacting components must be demonstrated to be biocompatible.	Achieved: Inner headband (Polyester, polyurethane) and transducer covers (medical grade silicone rubber) listed as materials with a long history of safe use in medical devices that contact intact skin, posing very low biocompatibility risk. No exclusion characteristics apply.
Electromagnetic Compatibility & Safety	Device must demonstrate electromagnetic compatibility, battery safety, and electrical safety.	Achieved: EMC tests (IEC 60601-1-2, 47 CFR FCC Part 15 Subpart B) conducted by SGS Taiwan Ltd. showed Sonu is electromagnetically compatible, electrically safe, and has a safe battery.
Software Performance	Software verification, validation, and hazard analysis must be performed.	Achieved: Documentation provided for Software Level of Concern (Minor), description, hazard analysis, requirements specification, traceability, V&V, and revision history. Cybersecurity assessment also conducted.
Non-clinical Performance	Device must perform as intended under anticipated conditions of use, including verification of specified mechanical stimulation parameters.	Achieved: 100% of 60 manufactured units passed battery and functional pairing tests. Audio quality and volume changes at max and 50% settings were verified. Therapeutic resonance measurement was verified.
Risk Mitigation (General)	Mitigation of risks such as injury from mechanical overstimulation (skin irritation, pain, discomfort, headache, vertigo, hearing loss, tinnitus) and ineffective treatment.	Achieved: Mitigation measures include non-clinical performance testing, human factors testing, software V&V and hazard analysis, electrical safety testing, EMC testing, battery safety testing, and labeling.
Human Factors	Human factors testing must demonstrate that users can successfully use the device in the intended use environment based solely on its labeling and instructions for use.	Achieved: Usability study with 25 participants (22+ years old) for OTC treatment of moderate to severe nasal congestion. 100% strongly agreed or agreed that instructions were easy to follow and the device was easy to use. No neutral/negative responses or observations of difficulty by the observer.

2. Sample Size and Data Provenance for the Test Set

Sample Size (Clinical Study - SCORE): 52 subjects (26 Sonu Therapy, 26 Sham).
Data Provenance: Prospective, multi-center, randomized, double-blinded, sham-controlled interventional study. The country of origin of the data is not explicitly stated, but the mention of a US-based sponsor ("Sound Health Systems, Inc.") and the FDA De Novo request implies the study was likely conducted in the US or under US regulatory oversight.

3. Number of Experts and their Qualifications for Ground Truth of the Test Set

The provided text describes a clinical trial where patient-reported outcomes were used as the primary and secondary endpoints.

0 Experts: No external experts were used to establish the "ground truth" for the test set. Instead, the ground truth was based on the subjects' self-reported symptom scores using a validated scoring system (TNSS).

4. Adjudication Method for the Test Set

None (implicitly): Since the primary and secondary outcomes were patient-reported self-assessments (nasal congestion sub-score and TNSS), there was no mention of an independent adjudication method for these scores. The blinding of both subjects and potentially study staff (though not explicitly stated for staff beyond the "double-blinded" claim) would have been the main control against bias.

5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study

No, this was not an MRMC study. The study evaluated the direct effectiveness of the device itself (Sonu Therapy vs. Sham) and did not involve human readers interpreting imaging or other data with and without AI assistance. Therefore, there is no effect size of human readers improving with AI vs. without AI.

6. Standalone (Algorithm Only) Performance Study

No, a standalone (algorithm only) performance study was not explicitly conducted or reported for the primary effectiveness of decongestion. The device functions as a complete system (Sonu Band + App + Cloud Engine) that delivers treatment to the human user.
The "Sonu Cloud Engine" does utilize an AI/ML algorithm to estimate patient-specific resonant frequencies based on 3D face scans. A preliminary version of the device ("Soniflow System") was evaluated in a human cadaver model to verify the resonance frequency calculation algorithm. This could be considered a form of standalone performance evaluation for that specific algorithmic component, but not for the overall decongestion claim. The text does not detail the acceptance criteria or results for this cadaver study beyond stating it was done for "verification."

7. Type of Ground Truth Used

Patient-Reported Outcomes (via validated scoring system): The primary and secondary effectiveness endpoints relied on subjects' self-reported nasal congestion sub-scores and Total Nasal Symptom Scores (TNSS). The TNSS is described as a "validated symptom severity scoring system."

8. Sample Size for the Training Set

Not explicitly stated for the AI/ML algorithm. The text mentions the Sonu Cloud Engine uses an "Artificial Intelligence / Machine Learning algorithm that has been trained to correlate craniofacial measurements with sinus dimensions." However, the size and nature of this training dataset are not provided.
For the clinical effectiveness study (SCORE), no dedicated "training set" in the machine learning sense is applicable, as this was an interventional clinical trial evaluating the device in humans, not training an AI for diagnosis or treatment.

9. How the Ground Truth for the Training Set was Established

Not explicitly stated for the AI/ML algorithm's training set. For the AI/ML algorithm that estimates patient-specific resonant frequencies, the text implies it was trained to "correlate craniofacial measurements with sinus dimensions." This would logically require a dataset where both craniofacial measurements (presumably from 3D scans) and actual sinus dimensions (e.g., from CT scans or other anatomical imaging) were available. However, the method for establishing this "ground truth" (e.g., expert anatomical measurements, verified imaging data) for the AI training is not described in the document.

Summary

Regulation Number and Section

N/A