Search Results

Hura CTP™ v1.0 is intended as an image processing software to reduce noise of head CT Perfusion (CTP) DICOM images through multiple algorithm steps.

The software reduces image noise and enhances image contrast (e.g. contrast-to-noise ratio (CNR) and signal-to-noise ratio (SNR)) of the CTP DICOM images. Hura CTP™ v1.0 is non-iterative; hence the low computational overhead enables fast processing and allows no interruption to clinical workflow.

Hura CTP™ v1.0 outputs head CTP DICOM images with enhanced image quality to a designated directory defined by the user. The processed DICOM images can be imported to a third-party post-processing software for quantification of hemodynamic parameters.

The use of this algorithm may enhance the image contrast of head CTP DICOM images depending on the clinical task, patient size, and clinical practice. A consultation with a radiologist and a physicist should be made to determine the appropriate imaging protocol to obtain diagnostic image quality for the clinical task.

Hura CTP™ v1.0 is intended for use only by trained and qualified clinical personnel (e.g. radiologists). Hura CTP™ v1.0 is also intended to be used by trained and qualified personnel for installation and maintenance of the software.

Hura CTP™ v1.0 is an image processing software which reduces noise of CTP DICOM images and enhances image contrast and signal-to-noise ratio. Hura CTP™ v1.0 is based on a new algorithm termed k-space weighted image average (KWIA) that was adapted from accelerated 4D dynamic MRI with projection view-sharing. There are two major advantages of KWIA compared to existing denoising method for CTP:

• 1. KWIA is computationally simple and fast (non-iterative); hence the low computation overhead enables fast processing and allows no interruption to clinical workflow.
• 2. KWIA does not make assumptions of noise characteristics and preserves the texture and resolution of CT images.

The software consists of three modules, namely the image input module, the processing module, and the output module. The image input module is responsible for interfacing with DICOM compliant CT scanners and receiving DICOM images. The image processing module is responsible for motion compensation, performing Fourier transform on DICOM images, applying KWIA, and performing inverse Fourier transform to output noise-reduced images. Both original and the noise-reduced DICOM images are then saved to the specified file directory. Hura CTP™ v1.0 is written in C/C++ language and runs as a local application on a standard PC, Mac, or UNIX workstation.

Insight Toolkit (ITK) serves as an important off-the-shelf library that KWIA algorithm leverages for a number of computational operations. The output module is responsible for transmitting noise-reduced CTP DICOM images to a designated directory defined by the user. The DICOM images can be imported to a third-party post-processing software (e.g. iNtuition, RAPID, Vitrea, etc.) for quantification of hemodynamic parameters. The software should be used only by trained professionals including, but not limited to, physicians, medical physicists, and technicians.

The provided text describes the acceptance criteria and a study proving the device meets these criteria. Here is a breakdown of the requested information:

Acceptance Criteria and Device Performance

The provided document does not explicitly state a table of "acceptance criteria" with defined thresholds that the device had to meet (e.g., "SNR increase of at least X%"). Instead, it presents performance metrics observed in the validation studies and refers to "significant reduction" or "significant increase" as evidence of effectiveness. The implicit acceptance criterion appears to be a statistically significant improvement in image quality metrics (SNR, CNR, SD reduction) and preservation of diagnostic information (low NRMSE of TDCs, excellent ICC for perfusion parameters).

Table of Acceptance Criteria and Reported Device Performance

Study Details

1. Sample Size Used for the Test Set and Data Provenance:

   • Phantom Validation Study: 30 CTP phantom scans. Data acquired on Philips Brilliance and Toshiba Aquilion scanners. Provenance is not specified (e.g., country of origin), but it's a phantom study.
   • Clinical Cases with Simulated Small Features: 40 CTP datasets of two clinical cases with inserted simulated small objects. Acquired on Siemens Sensation and Toshiba Aquilion scanners. Provenance is not specified.
   • Retrospective Clinical Study: 40 datasets of clinical DICOM images acquired at 4 sites on CT scanners manufactured by 4 OEMs (GE, Philips, Siemens, Toshiba). The data provenance describes it as a "retrospective clinical study," implying existing data, but the country of origin is not specified.

2. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of those Experts:
   The document does not mention the use of human experts to establish ground truth for the test set in the conventional sense (e.g., radiologists reviewing images and making diagnoses). The ground truth for the quantitative metrics (SNR, CNR, SD, NRMSE, ICC) appears to be derived from comparing the processed images (by Hura CTP™ v1.0) against the original vendor-generated images, and in the phantom study, likely against known phantom properties. The reference to "iNtuition" (predicate device) and "vendor generated DICOM images" suggests these serve as reference points for comparison rather than independent "expert ground truth" for evaluation.

3. Adjudication Method for the Test Set:
   Not applicable, as the evaluation methods described involve quantitative image metrics and software comparisons, not human reader adjudication of diagnostic findings.

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 involving human readers and AI assistance was not performed or described in this document. The study focuses purely on the technical performance of the image processing software in improving image quality metrics.

5. If a Standalone (i.e., algorithm only without human-in-the-loop performance) was done:
   Yes, the studies described are standalone performance evaluations of the Hura CTP™ v1.0 algorithm. The metrics (SNR, CNR, SD, NRMSE, ICC) are calculated directly from the image data processed by the algorithm without human interpretation or interaction.

6. The Type of Ground Truth Used (expert consensus, pathology, outcomes data, etc.):

   • Phantom Studies: The "ground truth" is implied by the known characteristics of the phantom and the comparison against original, vendor-generated images. The NRMSE and ICC calculations further evaluate the preservation of the original signal characteristics.
   • Clinical Studies (with simulated objects and retrospective): The "ground truth" for evaluation is based on comparing "Hura CTP™ v1.0 processed" images against "vendor generated DICOM images." The intent is to show that the algorithm improves objective image quality metrics (SNR, CNR, SD) while preserving the underlying signal/perfusion information (low NRMSE of TDCs, excellent ICC for perfusion parameters). It leverages a predicate device (iNtuition) for performance comparison, which suggests iNtuition's output is a "reference" for perfusion map generation.

7. The Sample Size for the Training Set:
   The document does not explicitly state the sample size of a "training set" for the Hura CTP™ v1.0 algorithm. It describes the algorithm (KWIA) and its characteristics but does not detail a machine learning training paradigm often associated with "training sets." If the algorithm is rule-based or deterministic, a training set might not be applicable in the same way as for deep learning models.

8. How the Ground Truth for the Training Set Was Established:
   Since no training set is mentioned in the context of the algorithm's development (only validation), this information is not provided. The algorithm is described as "adapted from accelerated 4D dynamic MRI with projection view-sharing," suggesting a more theoretical or model-based development rather than a data-driven training process with established ground truth labels.

Ask a specific question about this device