Search Results

AUTION EYE AI-4510 Urine Particle Analysis System is a fully automated urine particle analyzer for in vitro diagnostic use. AUTION EYE AI-4510 is intended for the quantitative measurement of red blood cells (WBC) and squamous epithelial cells (SQEC), the semi-quantitative measurement of bacteria (BACT) and crystals (CRYS) and the qualitative measurement of white blood cell clumps (WBCC), non-squamous epithelial cells (NSE), hyaline casts (HYAL), non-hyaline casts (NHC), yeast (YST), mucus (MUCS) and sperm (SPRM) in urine samples.

A trained operator can set criteria for flagging speciment analyte image decisions should be reviewed and reclassified as necessary by a trained technologist.

The AUTION EYE AI-4510 analyzer can be used as a standalone unit or combined with an AUTION MAX AX-4060 urine chemistry analyzer.

The AI-4510 System (AUTION EYE AI-4510) is a fully automated urine particle analyzer for in vitro diagnostic use that uses flow cell digital imaging technology in a clinical laboratory setting. Based on images captured in the flow method, the instrument automatically classifies the images of various formed elements. The AI-4510 System can quantitatively measure RBC, WBC, and SQEC; semi-quantitatively measure BACT, and CRYS; and qualitatively measure WBCC, NSE, HYAL, NHC, YST, MUCS and SPRM in urine samples. In addition, the AI-4510 System allows trained operators to manually review and reclassify all the element images collected by the system.

This document describes the validation of the AUTION EYE AI-4510 Urine Particle Analysis System. The device is intended for the quantitative, semi-quantitative, and qualitative measurement of various elements in urine samples. The validation primarily focuses on demonstrating substantial equivalence to a legally marketed predicate device (iQ200 Urine Analyzer).

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

1. Acceptance Criteria and Reported Device Performance

The document doesn't explicitly list "acceptance criteria" in a single table for all performance measures. Instead, it states that "all results meeting the predefined acceptance criteria" for precision studies, and that "Quantitative, Semiquantitative and Qualitative parameters met the acceptance criteria" for the method comparison study. The reported performance is presented in various tables throughout the "Summary of Performance Data" section (5.6).

Here's an aggregated table derived from the provided performance data:

2. Sample size used for the test set and the data provenance

• Precision Studies:
  • Repeatability Study: Clinical urine samples were used.
    • Quantitative elements: Not explicitly stated, but "clinical urine samples in the evaluation of repeatability for all twelve (12) elements from low to high concentrations."
    • Semi-quantitative and Qualitative elements: n=10 replicates per test level (e.g., Level 1, 2, 3 etc.).
  • Within-Laboratory Precision Study: ARKRAY control materials prepared using clinical samples.
  • Reproducibility Study: Commercially available control materials and ARKRAY control materials prepared using clinical samples were used.
• Linearity Testing: Not specified for sample size beyond "one instrument."
• Limit of Detection: Not specified for sample size.
• Carryover Testing: High-level and low-level samples, aliquoted into 5 tubes each, measured in sequences (e.g., H1 L1 H2 L2 H3 L3 H4 L4 H5 L5, repeated 5 times).
• Interference Testing: Not specified for sample size beyond the substances tested.
• Sample Stability: Positive and negative samples for all 12 elements.
• Method Comparison:
  • Population for Reference Range results: n=247
  • Quantitative Elements (RBC, WBC, SQEC): n=377 (RBC), n=845 (WBC), n=382 (SQEC) for comparison between AI-4510 (M) and iQ200 (M).
  • Semi-quantitative & Qualitative Elements: n=1474 (CRYS, BACT, WBCC, MUCS, SPRM), n=765 (NSE, NHC, HYAL, YST).
  • Data Provenance: Clinical samples. The method comparison study was conducted at "three (3) CLIA-Moderate complexity laboratories." The document states samples were "collected fresh within two (2) hours or refrigerated up to six (6) hours post collection," implying a prospective collection directly for these studies. The country of origin is not explicitly stated, but the submission is for FDA clearance in the US, and the company and testing sites (CLIA labs) suggest operations relevant to the US market.

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

The ground truth for the method comparison study (especially for semi-quantitative and qualitative elements) appears to be established by comparison to the predicate device iQ200 (Manual) and Manual Microscopy.

• For the method comparison, it refers to "CLIA-trained operators" performing testing and "A trained operator can set criteria for flagging speciment analyte image decisions should be reviewed and reclassified as necessary by a trained technologist." This indicates that trained technologists or CLIA-trained operators (which implies suitable qualifications for laboratory testing) established the "ground truth" or reference values, either by manual microscopy or using the predicate device's manual review function.
• The document does not specify the exact number of individual experts or their specific qualifications (e.g., specific years of experience, board certification as pathologists or medical technologists). It only refers to "CLIA-trained operators" and "trained technologists."

4. Adjudication method for the test set

• The document states: "All instrument analyte image decisions should be reviewed and reclassified as necessary by a trained technologist." This implies a form of human override or adjudication post-AI classification.
• For the "Method Comparison" tables (14, 15, 16), most comparisons are listed as "AI-4510 (Manual) vs. iQ200 (Manual)" or "AI-4510 (Manual) vs. Manual Microscopy." The "(M)" denotes "manually reviewed and reclassified results." This indicates that the results from both the investigational device and the predicate device/manual microscopy were subjected to manual review/adjudication by trained human operators to establish the final classification used for comparison.
• The specific method of adjudication (e.g., 2+1, 3+1 consensus) among multiple readers for establishing the ground truth is not specified. The comparison is against already "manual" classifications from the predicate or direct manual microscopy, suggesting that the human reading itself serves as the reference, likely by one or more trained technologists.

5. 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

• This document describes performance characteristics of the device itself, often compared to a predicate device or manual microscopy.
• It does not describe an MRMC comparative effectiveness study where the performance of human readers with AI assistance is directly compared to human readers without AI assistance to quantify improvement or effect size. The AI-4510 System is an automated analyzer with a manual review component, not an AI assistance tool for human interpretation of images outside of the system.

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

• The document mentions "The software processes the recorded images, automatically identifying and classifying the formed elements based on the sorting algorithm." (Section 5.5)
• However, the indications for use explicitly state: "A trained operator can set criteria for flagging specimens. All instrument analyte image decisions should be reviewed and reclassified as necessary by a trained technologist."
• Furthermore, the "Method Comparison" tables are predominantly listed as "AI-4510 (Manual) vs. iQ200 (Manual)" or "AI-4510 (Manual) vs. Manual Microscopy," where "(M)" denotes "manually reviewed and reclassified results."
• This strongly suggests that the reported performance data for clinical claims (method comparison) represents the combined human-in-the-loop performance after technologist review and reclassification, particularly for the semi-quantitative and qualitative elements.
• While the device has an "automatic classification" function (also mentioned in section 5.4 under "Automatic Classification"), the reported clinical performance data does not appear to be purely standalone (algorithm-only) without human intervention.
• Table 11 (Interference Effect on Auto-classified Results) hints at some testing of the auto-classified performance in specific scenarios (interference), but the bulk of the clinical validation on the main intended use appears to involve human review.

7. The type of ground truth used

The ground truth used for the method comparison study was established through:

• Comparison to the iQ200 System (Manual): This means the results obtained from the predicate device after its own manual review and reclassification process.
• Manual Microscopy: This is considered the traditional gold standard for urine particle analysis, established by trained technologists.

Therefore, the ground truth is a combination of expert consensus (implied via "trained technologist" review) and comparison to a legally marketed predicate device (also with human review), with manual microscopy serving as a reference.

8. The sample size for the training set

The document provided does not contain any information about the training set for the AI-4510 System's algorithm. This K submission focuses on device performance studies for validation and comparison to a predicate, not on the developmental aspects of the AI model.

9. How the ground truth for the training set was established

As no information about the training set is provided, how its ground truth was established is also not available in this document.

Ask a specific question about this device