(67 days)
UROSTATION - 3D PROSTATE SUITE with MRI/3DTRUS fusion option and with Second Look 3DTRUS fusion option is a computer-based software application intended to process, visualize and record 3D digital ultrasound images of the prostate.
UROSTATION - 3D PROSTATE SUITE with MRI/3DTRUS fusion option and with Second Look 3DTRUS fusion option is intended to be used by physicians in the clinic or hospital for 2D and 3D visualization of the prostate gland and for the 3D transrectal ultrasound based fusion of multiple imaging modalities (ultrasound, MRI) in order to map such prostate gland. Additional software features include patient data management, multimodal data communication, multiplanar reconstruction, surface and volume rendering, organ delineation, region of interest delineation, 3D image registration and data reporting.
UROSTATION - 3D PROSTATE SUITE is a computer-based software application designed to process, visualize and record 3D digital ultrasound images of the prostate, and to manage patient and clinical data in the context of transrectal prostate biopsy. Software options provide the fusion of 3DTRUS with MRI or with other 3DTRUS modalities.
Hardware Platform and Operating System
The application runs on a standard Personal Computer under Microsoft Windows® operating system (version 7 or higher).
Peripheral and accessories
The application is controlled by a footswitch and manual input devices (mousc, kevboard),
It is designed to work in connection with commercially available 3D ultrasound scanners with Ethernet connection, 3D transrectal ultrasound probe and needle guide.
Software Features
UROSTATION - 3D PROSTATE SUITE implements image fusion and display algorithms to provide 3D representation of prostate biopsies.
A typical workflow enables the physician to intraoperatively visualize the 3D mapping of biopsies with respect to a 3DTRUS reference image of the patient's prostate.
For that purpose. 3D digital images may be transferred at any time from the 3D ultrasound scanner to the Urostation for registration and display, while the physician keeps track of the organ on the ultrasound scanner using the usual 2D live B mode.
Optionally, MRI/3DTRUS fusion allows the elastic registration of the 3DTRUS reference image with other imaging modalities (MRI here) in order to display the 3D biopsy mapping on multiple imaging modalities.
Optionally, Second Look 3DTRUS Fusion allows the elastic registration of the 3DTRUS reference image with a previously acquired 3DTRUS reference image of the same patient in order to superimpose two 3D biopsy mappings on a unique 3DTRUS reference image.
Alternatively, UROSTATION - 3D PROSTATE SUITE also provides a review mode that allows the mapping of histologic results on the said 3D reference image of the patient's prostate. Patient information, images and 3D biopsy mapping may be stored or printed for future retrieval and examination.
The acceptance criteria for the UROSTATION - 3D PROSTATE SUITE with MRI/3DTRUS fusion option and Second Look 3DTRUS fusion option are based on its performance and accuracy in image fusion and organ tracking. The device aims to assist physicians in targeting different regions of the prostate by registering biopsy cores and optionally MR lesions on a reference 3DTRUS image.
The studies presented validate the performance and accuracy of the KOELIS fusion technology on both phantom and patient data.
Acceptance Criteria and Reported Device Performance
| Acceptance Criteria Category | Specific Metric | Acceptance Criteria (Implied by conclusion) | Reported Device Performance |
|---|---|---|---|
| Elastic MR/3DTRUS Image Fusion | Mean distance between reference and deformed prostates (after elastic registration) | Low error, indicating accurate registration of prostate contours | 0.09 mm |
| RMS error stemming from deformations in the probe area | Reduced error compared to pre-registration | Reduced from 4.1 mm down to 1.9 mm | |
| Mean accuracy of superimposed surfaces after elastic registration | High accuracy in surface superposition | 0.7 ± 0.3 mm | |
| Mean distance between corresponding landmarks after elastic registration | Low error in landmark registration | 1.7 ± 0.7 mm | |
| RMS error for landmark registration | Low error | 1.9 mm | |
| Elastic 3DTRUS/3DTRUS Image Fusion & Organ Tracking | System robustness (registration failures) | Low percentage of registration failures | 17 failures out of 786 (2%) biopsy volumes |
| Accuracy of organ tracking using fiducials | Low error in tracking prostate tissue motion | 0.76 mm ± 0.52 mm | |
| Targeting Accuracy (Hypoechoic Lesions) | Percentage of successful hits for hypoechoic lesions | High success rate | 100% (27 out of 27 biopsies) |
| Procedural targeting error (hypoechoic lesions) | Low error | 1.52 ± 0.78 mm | |
| System registration error (hypoechoic lesions) | Low error | 0.83 mm | |
| Overall error (hypoechoic lesions) | Low overall error for targeting | 2.35 mm | |
| Targeting Accuracy (Isoechoic Lesions with MR Fusion) | Percentage of successful hits for isoechoic lesions | High success rate | 84% (24 out of 27 MR fusion biopsies) |
| Mean procedural targeting error (isoechoic lesions) | Low error | 2.09 ± 1.28 mm | |
| Overall error (isoechoic lesions) | Low overall error for targeting | 2.92 mm | |
| Overall Visualization Process | Total RMS error of the visualization process | Clinically acceptable total RMS error, leading to high hit probability | 2.05 mm (combining 1.9mm for MRI/3DTRUS fusion and 0.76mm for 3DTRUS/3DTRUS fusion). This gives a probability greater than 95% to hit the lesion according to Karnick et al. Med. Phys. 37 (2), 802-813. KOELIS concluded this accuracy is clinically acceptable. |
Study Details:
-
Sample sizes used for the test set and the data provenance:
- Elastic MR/3DTRUS image fusion (Brolis et al, 2012):
- Test set: 17 contoured TRUS patient volumes for simulated deformations, resulting in 850 simulated deformations.
- Clinical Validation: 49 patients.
- Data Provenance: Not explicitly stated, but "5 clinical sites" suggests multi-center data, likely prospective clinical data given the "clinical validation" and "landmark pairs were identified and approved by physicians".
- Elastic 3DTRUS/3DTRUS image fusion and organ tracking (Baumann et al, 2012):
- Test set: 786 biopsy volumes acquired from 47 patients.
- Accuracy evaluation: 687 registered volumes stemming from 40 patients.
- Data Provenance: Not explicitly stated, but implies prospective data ("during biopsy sessions").
- Targeting Accuracy of Urostation (Ukimura et al, 2012):
- Test set: 6 prostate phantoms (3 containing 3 hypoechoic lesions and 3 containing 3 isoechoic but MRI-visible lesions).
- Data Provenance: Phantoms (simulated environment). This study also references "clinical validation" in other sections, but for targeting accuracy, it's explicitly phantom-based.
- Elastic MR/3DTRUS image fusion (Brolis et al, 2012):
-
Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
- Elastic MR/3DTRUS image fusion (Brolis et al, 2012): "Physicians" identified and approved 112 landmark pairs. Specific number and qualifications beyond "physicians" are not provided.
- Elastic 3DTRUS/3DTRUS image fusion and organ tracking (Baumann et al, 2012): Not explicitly stated how ground truth for accuracy evaluation (using fiducials) was established, but implies fiducial markers as objective ground truth.
- Targeting Accuracy of Urostation (Ukimura et al, 2012): The ground truth for lesion location in phantoms is inherent to the phantom design. The accuracy of biopsy "hitting the lesion" is a direct measurement based on the physical phantom. No external experts are mentioned for ground truth establishment for this specific study.
-
Adjudication method (e.g. 2+1, 3+1, none) for the test set:
- Not explicitly stated for any of the studies. For the clinical validation in Brolis et al., "landmark pairs were identified and approved by physicians," which suggests a consensus or verification process, but the specific method is not detailed.
-
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:
- The provided information does not include an MRMC comparative effectiveness study that assesses the improvement of human readers with AI assistance versus without. The studies focus on the intrinsic accuracy and performance of the fusion technology itself.
-
If a standalone (i.e. algorithm only without human-in-the-loop performance) was done:
- Yes, the "Performance summary" section primarily describes standalone algorithm performance in terms of registration accuracy, error rates, and hit probabilities for both fusion capabilities and targeting accuracy. The Brolis et al. study on "simulated deformations" and the Baumann et al. study on "system robustenss" and "accuracy was evaluated to 0.76mm±0.52mm using fiducials" directly assesses the algorithm's performance. Similarly, the phantom study by Ukimura et al. measures the algorithmic accuracy of targeting.
-
The type of ground truth used (expert consensus, pathology, outcomes data, etc.):
- Brolis et al. (Elastic MR/3DTRUS): Expert-identified landmark pairs and contours (from 3DTRUS patient volumes) for registration accuracy.
- Baumann et al. (Elastic 3DTRUS/3DTRUS): Fiducials (objective physical markers) for evaluating tracking accuracy.
- Ukimura et al. (Targeting Accuracy): Physical lesion locations within prostate phantoms. Biopsy success was likely determined by confirming needle placement relative to the known lesion locations in the phantoms.
-
The sample size for the training set:
- The document does not explicitly state the sample size for the training set for any of the algorithms. These studies primarily describe validation/testing.
-
How the ground truth for the training set was established:
- Since the training set size and details are not provided, how its ground truth was established is also not detailed in this document.
§ 892.2050 Medical image management and processing system.
(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).