Search Results

LiverSmart is indicated to:

For Liver Iron Concentration

1. measure liver iron concentration in individuals with confirmed or suspected systemic iron overload;

2. monitor liver iron burden in transfusion dependents and patients with sickle cell disease receiving blood transfusions:

3. aid in the identification and monitoring of non-transfusion-dependent thalassemia patients receiving therapy with Deferasirox.

For Liver Fat Assessment

1. assess the volumetric liver fat fraction, proton density fat fraction and steatosis grade in individuals with confirmed or suspected fatty liver disease.

When interpreted by a trained physician, the results can be used to:

2. monitor liver fat content in patients undergoing weight loss management;

3. aid in the assessment and screening of living donors for liver transplant.

LiverSmart is software that utilizes two existing FDA cleared devices, HepaFat-AI (K201039) and FerriSmart (K182218) and combines their respective results into a singular consolidated multiparametric 'LiverSmart' report.

LiverSmart automatically sorts and sends magnetic resonance imaging (MRI) datasets to each of the existing HepaFat-AI and FerriSmart devices and then receives results from those devices to generate a summary report which combines the HepaFat-AI results (an estimate of the patient's volumetric liver fat fraction (VLFF), proton density fat fraction (PDFF), steatosis grade), and the FerriSmat result (an estimate of the patient's liver iron concentration (LIC)).

To conduct analysis, the user simply uploads a single zipped folder containing HepaFat-AI and FerriSmart DICOM images, acquired in accordance with their respective acquisition protocols, to the LiverSmart software. No user input is required for the analysis thereby minimising the impact of human error. The LiverSmart software requires image input data that has been acquired in accordance with the existing and now well established HepaFat-AI (K201039) and FerriSmart (K182218) imaging protocols.

LiverSmart has two new components that are in addition to the existing components of HepaFat-AI and FerriSmart, namely a:

• (i) Data Preparation Module; and
• (ii) Report Generation Module

The rest of the components for LiverSmart are the existing components of the FDA cleared HepaFat-AI and FerriSmart devices, as follows:

For HepaFat-AI:

• (i) Magnetic Resonance Imaging Protocol
• (ii) HepaFat-AI Analysis Software
• (iii) Volumetric Liver Fat Fraction Measurement
• (iv) Proton Density Fat Fraction Measurement
• (v) Steatosis Grade Measurement

For FerriSmart:

• (i) Magnetic Resonance Imaging Protocol
• (ii) FerriSmart Analysis Software
• (iii) Liver Iron Concentration Measurement

The above HepaFat-AI and FerriSmart components are the same as previously provided to the FDA as the time HepaFat-AI and FerriSmart regulatory clearances were sought (and subsequently obtained).

The provided document, K213776, describes the LiverSmart device, which combines the functionalities of two previously cleared devices, FerriSmart (K182218) and HepaFat-AI (K201039). The core claim is that LiverSmart is substantially equivalent to these predicates, not that it offers improved performance beyond what they individually provide. Therefore, the "acceptance criteria" and "study that proves the device meets the acceptance criteria" in this context refer to demonstrating that LiverSmart accurately integrates the functions of its predicates and produces identical results from the same input data, as opposed to proving novel clinical performance.

Here's an analysis based on the provided text:

Acceptance Criteria and Reported Device Performance

The acceptance criteria for LiverSmart are implicitly derived from the established performance of its predicate devices, FerriSmart and HepaFat-AI, and the requirement for LiverSmart to accurately integrate and reproduce their results.

Study Proving Acceptance Criteria (Verification Testing)

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

• The document states "Verification testing confirms that the data preparation module of LiverSmart detects anomalies..." and "Additionally, LiverSmart yields identical results...when the same image datasets are analysed by HepaFat-AI and FerriSmart devices independently."
• Sample Size: The exact sample size used for this verification testing (test set) is not explicitly stated in the provided text.
• Data Provenance: The document does not specify the country of origin of the data or whether it was retrospective or prospective. It only mentions using "the same image datasets" as analyzed by the predicate devices.

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

• This question is not applicable in the traditional sense for this specific 510(k) submission. The LiverSmart device is an integration of two already cleared devices. The "ground truth" for the individual measurements (LIC, VLFF, PDFF, steatosis grade) would have been established during the original clearance processes for FerriSmart and HepaFat-AI, likely through comparison with biopsy or other approved methods.
• For LiverSmart, the verification testing focuses on the concordance of results between LiverSmart and its predicates, not on establishing new clinical ground truth for the measurements themselves. Therefore, no new "experts" are noted as establishing ground truth for the LiverSmart test set in this document. The "trained physician" mentioned in the Indications for Use is for interpretation, not for establishing algorithmic ground truth.

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

• The document does not specify any adjudication method for the verification testing. The primary "adjudication" is the direct comparison of LiverSmart's output to the output of FerriSmart and HepaFat-AI on the same datasets, aiming for "identical results."

5. If a multi-reader multi-case (MRMC) comparative effectiveness study was done:

• An MRMC comparative effectiveness study was not conducted for LiverSmart as described in this document. LiverSmart is presented as an integration of existing cleared devices, not an improvement or replacement necessitating a comparative effectiveness study involving human readers' diagnostic performance. The document focuses on the algorithmic output matching previously cleared algorithms.

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

• Yes, the verification testing described for LiverSmart is a standalone (algorithm only) performance assessment. The text states LiverSmart "automatically sorts and sends...receives results...to generate a summary report" and "No user input is required for the analysis thereby minimising the impact of human error." The verification confirms that LiverSmart's algorithmic output is identical to that of its predicate algorithms on the same input.

7. The type of ground truth used:

• For the verification of LiverSmart, the "ground truth" is effectively the results produced by the FDA-cleared predicate devices (FerriSmart and HepaFat-AI) when processing the same image datasets. The goal was to demonstrate "identical results."
• The original ground truth methodologies (e.g., expert consensus, pathology, outcomes data) for the measurements themselves (LIC, VLFF, PDFF, steatosis grade) would have been established and reviewed during the original 510(k) clearances for FerriSmart and HepaFat-AI, but are not detailed here for LiverSmart.

8. The sample size for the training set:

• The document states that LiverSmart utilizes the "existing components of the FDA cleared HepaFat-AI and FerriSmart devices." It further notes that "The above HepaFat-AI and FerriSmart components are the same as previously provided to the FDA at the time HepaFat-AI and FerriSmart regulatory clearances were sought (and subsequently obtained)."
• Therefore, the training sets (and their sizes) for the underlying convolutional neural networks (CNNs) would pertain to the development of FerriSmart and HepaFat-AI. The sample size for these training sets is not provided in this document (K213776).

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

• Similar to point 8, the ground truth for the training sets would have been established during the development and clearance of FerriSmart and HepaFat-AI.
• The document mentions "Convolutional neural networks for the image analysis" for both LiverSmart and its predicates, implying that these models were trained. However, this document (K213776) does not detail how the ground truth for the training sets of the predicate devices was established. This information would be found in the original 510(k) submissions for FerriSmart (K182218) and HepaFat-AI (K201039).

Ask a specific question about this device

FerriSmart is indicated to:
· measure liver iron concentration in individuals with confirmed or suspected systemic iron overload;
· monitor liver iron burden in transfusion dependent thalassemia patients with sickle cell disease receiving blood transfusions:
• aid in the identification and monitoring of non-transfusion-dependent thalassemia patients receiving therapy with deferasirox.

FerriSmart is a stand-alone software application that automatically analyses multi-slice, spin-echo MRI data sets encompassing the abdomen to determine the signal decay rate (R>) that is used to characterize iron loading in the liver, which is then transformed by a defined calibration curve to provide a quantitative measure of liver iron concentrations in vivo.

The software application is a measuring medical device intended to be hosted either in a cloudbased or on site hosted platform and used directly by the radiographer. It does not drive the MRI machine and does not come into direct contact with patients.

The key components of FerriSmart are:

• Specific Magnetic Resonance Imaging Protocol: Use of a specific magnetic resonance imaging protocol for acquisition of the raw image data. The imaging protocol is critical to ensure the quality of the end results. Its adherence is verified by the IQC Module, an automated algorithm that checks the correctness of the parameters of the data acquisition protocol.
• FerriSmart AI Analysis Software: Custom-designed image analysis software performing the R2 measurement based on AI (Artificial Intelligence) technology.
• An additional software module (algorithmic) that Liver Iron Measurement: incorporates a calibration curve relating R2 to liver iron concentration (LIC) is added to allow production of a liver iron concentration report.

The result report provides the patient's average LIC reported in micromole and milligram per gram dry weight of liver. The images analysed are included in the report for review by the radiologist. The results are intended to assist in clinical diagnosis, and/or in making decisions concerning clinical management.

Here's a breakdown of the acceptance criteria and the study that proves the device meets them, based on the provided text:

Acceptance Criteria and Reported Device Performance

LIC Threshold: 3.0 mg Fe/g dry tissue (interrupt deferasirox for NTDT):
Sensitivity: 96% (95% CI 94-97%)
Specificity: 95% (95% CI 92-98%)

LIC Threshold: 3.2 mg Fe/g dry tissue (historical HHC definition, lower optimal for chelation):
Sensitivity: 94% (95% CI 92-96%)
Specificity: 95% (95% CI 92-98%)

LIC Threshold: 5.0 mg Fe/g dry tissue (consider deferasirox for NTDT):
Sensitivity: 91% (95% CI 89-94%)
Specificity: 97% (95% CI 95-99%)

LIC Threshold: 7.0 mg Fe/g dry tissue (upper optimal for chelation, increased risk):
Sensitivity: 92% (95% CI 90-95%)
Specificity: 97% (95% CI 95-98%)

LIC Threshold: 15.0 mg Fe/g dry tissue (greatly increased cardiac risk, increase deferasirox dose):
Sensitivity: 89% (95% CI 85-93%)
Specificity: 98% (95% CI 98-99%)

Overall, most sensitivities and specificities are above 90%, with the exception of specificity at 1.8 mg Fe/g dry tissue (80%) and sensitivity at 15.0 mg Fe/g dry tissue (89%). These exceptions are deemed acceptable for clinical use. |
| Usability | All participants found the product easy to use, fast, and technically reliable (no bugs). |
| Software Verification & Validation | Developed, verified, and validated following Design Control principles and General Principles of Software Validation guidelines. |

Study Information

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

   • Repeatability Study Test Set: 60 subjects scanned twice. The provenance of this data (country of origin, retrospective/prospective) is not specified.
   • Clinical Study Test Set: 971 datasets from multiple makes and models of MRI scanners. The provenance (country of origin, retrospective/prospective) is not specified.

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

   • The ground truth for the clinical study was the predicate device, FerriScan R2-MRI Analysis System. For the predicate, "human interaction for Region of Interest (ROI) selection" was noted.
   • The text does not specify the number or qualifications of experts involved in establishing the FerriScan results used as ground truth for this FerriSmart study. It only mentions that FerriScan is used "in-house by Resonance Health's analysts."

3. Adjudication method for the test set:

   • The text does not explicitly state an adjudication method (such as 2+1 or 3+1) for the comparison between FerriSmart and FerriScan results, or for the FerriScan results themselves. The ground truth was based on the FerriScan device output.

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:

   • No, an MRMC comparative effectiveness study was not done. The study's clinical performance evaluation was a standalone performance assessment of FerriSmart against a predicate device (FerriScan), not a comparison of human readers with vs. without AI assistance. The user of FerriSmart is stated to be a radiologist, who oversees the report, but the study focuses on the algorithm's performance relative to the predicate.

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

   • Yes, a standalone performance study was done. FerriSmart is described as a "stand-alone software application" which "automatically analyses" MRI data. The clinical study assessed its R2 and LIC measurements and diagnostic performance against the predicate without direct human intervention in the analysis process besides the radiologist reviewing the final report. FerriSmart uses an algorithm for automatic quality checks, whereas the predicate "requires human input" for some checks.

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

   • The ground truth for the study was the results from the predicate device, FerriScan R2-MRI Analysis System. FerriScan itself uses an algorithmic approach with human interaction for ROI selection and aims to provide quantitative measures of LIC, which would ultimately correlate to other clinical ground truths like liver biopsy in its own validation studies (as hinted by the comparison to biopsy error rates).

7. The sample size for the training set:

   • The document states, "FerriSmart AI Analysis Software has been trained on FerriScan data." However, the sample size for the training set is not explicitly provided in the given text.

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

   • The ground truth for the FerriSmart training set was established using data processed by the predicate device, FerriScan R2-MRI Analysis System. The text explicitly states, "FerriSmart AI Analysis Software has been trained on FerriScan data." This implies that the outputs from FerriScan (R2 measurements and LIC values) were used as the target for the AI's learning process.

Ask a specific question about this device

The FerriScan R2-MRI Analysis System is intended to measure liver iron concentration to aid in the identification and monitoring of non-transfusiondependent thalassemia patients receiving therapy with deferasirox.

The FerriScan R2-MRI Analysis System is a post-processing software tool that measures liver iron concentration based on the proton transverse relaxation rate (R2) of MRI images. R2 values are converted to liver iron concentration measurements using a calibration curve.

The FerriScan R2-MRI Analysis System is a post-processing software tool that measures liver iron concentration (LIC). It is intended to aid in the identification and monitoring of non-transfusion-dependent thalassemia (NTDT) patients receiving deferasirox therapy.

Here's an analysis of the acceptance criteria and the studies proving the device meets them:

1. Table of Acceptance Criteria and Reported Device Performance

The provided text excerpts do not explicitly list formal "acceptance criteria" with numerical thresholds for performance metrics. Instead, the document describes the required performance testing and the results of those tests, along with risks and mitigation measures that implicitly define acceptable performance. The key performance indicators evaluated are: Precision, Bias, Repeatability, Reproducibility, Sensitivity, and Specificity.

Therefore, the table below consolidates the relevant performance data from the provided text, indicating how the device performed against the implicitly required characteristics.

• Validation Study (233 patients): Bland-Altman 95% limits of agreement: 74% and -71%. |
  | Bias (systematic measurement error) | - Calibration Study (105 patients): Bland-Altman 95% limits of agreement with liver biopsy: -56% to 50% with bias of -3%.
• Validation Study (233 patients): Bland-Altman 95% limits of agreement: 74% and -71% with a bias of 1.9%.
• Mean percentage differences in LIC between FerriScan and liver biopsy were not significantly different than zero in either study. |
  | Repeatability (precision under same conditions over short period) | - 60 individuals tested twice: Standard deviation in R2 measurement: 8.1%.
• Consistent with 7.7% random error from initial 10-patient calibration study. |
  | Reproducibility (precision under different locations/operators) | - Phantom testing (K043271): Coefficient of variability across 13 different scanners: < 2.1%. |
  | Sensitivity at various LIC thresholds (95% CI) | - 1.8 mg Fe/g dw: 94% (86-97)
• 3.2 mg Fe/g dw: 94% (85-98)
• 7.0 mg Fe/g dw: 89% (79-95)
• 15 mg Fe/g dw: 85% (70-94) |
  | Specificity at various LIC thresholds (95% CI) | - 1.8 mg Fe/g dw: 100% (88-100)
• 3.2 mg Fe/g dw: 100% (91-100)
• 7.0 mg Fe/g dw: 96% (86-99)
• 15 mg Fe/g dw: 89% (83-96) |
  | Acceptance testing of images prior to processing | Labeling specifies instructions (e.g., FerriScan Phantom Pack use, acquisition settings, visual inspection, motion assessment). |
  | Data processing quality assurance protocols | Labeling describes protocols for phantom and patient image processing (e.g., ROI selection, noise assessment, motion correction). |

2. Sample Sizes and Data Provenance

• Test Sets (Clinical Studies for Precision and Bias):
  • Calibration Study: 105 patients.
  • Validation Study (subgroup from ESCALATOR trial): 233 patients.
  • Repeatability Study: 60 individuals.
  • Sensitivity/Specificity Study (from K043271): No specific number provided for the sensitivity/specificity calculation itself, but likely derived from the calibration study population.
• Data Provenance: The document does not explicitly state the country of origin for the patient data for these specific performance studies. However, the contact for the device is in Australia, and the deferasirox trials mentioned were likely international. The studies described are clinical studies, implying prospective data collection focused on the device's performance in patient populations, although the sensitivity/specificity study dates back to the original 510(k) (K043271), making it potentially a retrospective analysis of previously collected data.

3. Number of Experts and Qualifications for Ground Truth

The document does not specify the number or qualifications of experts used to establish the ground truth for the test sets.

4. Adjudication Method for the Test Set

The document does not describe any expert adjudication method (e.g., 2+1, 3+1) for the test sets.

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

No Multi-Reader Multi-Case (MRMC) comparative effectiveness study evaluating human readers with vs. without AI assistance is mentioned. The device is a "standalone" image post-processing system that provides a numerical LIC value, not an AI-assisted interpretation by a human reader.

6. Standalone Performance (Algorithm Only without Human-in-the-Loop)

Yes, standalone performance was done. The FerriScan R2-MRI Analysis System is described as a "post-processing software tool that measures liver iron concentration." The performance metrics (precision, bias, repeatability, reproducibility, sensitivity, specificity) are inherently measures of the algorithm's direct output (LIC value) compared to the ground truth. There is no mention of "human-in-the-loop" interaction for interpreting the R2-MRI calculation itself.

7. Type of Ground Truth Used

The primary ground truth used for assessing the device's performance (precision, bias, sensitivity, and specificity) was:

• Atomic absorption spectrometry from liver biopsy: This is considered the reference measurement of LIC.

8. Sample Size for the Training Set

The document does not explicitly state the sample size used for training the FerriScan R2-MRI Analysis System. The calibration study (105 patients) was used to "define an empirically-derived relationship" between R2 values and LIC, which implies it contributed to the creation or refinement of the calibration curve within the software. It's possible this study served as or contributed to the training data.

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

Assuming the "calibration study" described (105 patients) contributed to establishing the relationship used within the software (effectively the "training" aspect for the calibration curve), the ground truth for this calibration was established using atomic absorption spectrometry from liver biopsy.

Ask a specific question about this device