Search Results

CT CoPilot™ is intended for automatic labeling, visualization and volumetric quantification of segmentable structures from sets of CT images of the brain. This software is intended to automate the current manual process of identifying, labeling and quantifying structures identified on CT images of the brain and to provide automated registration and reformatting of data.

CT CoPilot™ is intended for use in automating post-acquisition quantitative analysis of CT images of the brain for patients aged 18 or older. CT CoPilot™ performs automatic reformatting, labeling and quantification of segmentable structures from a set of CT images. Output of the software provides these values as numerical volumes and images which have been annotated with graphical color overlays, with each color representing a specific segmental structure. When CT imaging is performed more than once on a patient, the current data is coregistered to the most recent processed prior exam of the same patient, facilitating comparison between the studies using CT CoPilot™. Voxel-by-voxel subtraction maps of the pixel density change in Hounsfield Units (HU) are generated in up to 3 dimensions between the current and most recent processed prior exam of the patient.

CT CoPilot™ incorporates registration, alignment, and segmentation methods similar to a previous 510(k) cleared device, known as NeuroQuant (K061855), to automatically label and quantify the volume of segmentable structures in MRI images of the head. CT CoPilot™ output is provided in standard DICOM format as additional series of images (with appropriate descriptors) and reports that can be displayed on most third-party commercial DICOM workstations. CT CoPilot™ is intended to provide visualization and quantification data for CT scans of the brain. CT CoPilot™ includes safety procedures and error reporting similar to those adopted by NeuroQuant to identify cases that may not be processed for any reason. CT CoPilot™ is intended to be used by trained personnel in Neuro CT imaging. Patient management decisions should not be made based solely on the results of CT CoPilot™ quantitative data.

In laboratory testing, CT CoPilot™ demonstrates the following registration accuracy based on 100 randomly acquired CT head scans on both normal patients and those with abnormal pathologies: 1) The inter-subject variability of the angle formed by the inter-hemispheric plane and the vertical line, as measured on axial views (yaw), is less than 15 degrees. 2) The intersubject variability of the positioning of the inter-hemispheric plane and the vertical line, as measured on coronal views (roll), is less than 15 degrees. 3) The inter-subject variability of the positioning of the AC/PC plane and the horizontal line, as measured on sagittal views, is less than 15 degrees. Laboratory testing of CT CoPilot™ software on 179 scans from 34 patients with ventriculostomy catheters demonstrates the following correlation coefficients between the automatic software segmentation accuracy of relevant anatomical structures when compared against the same group of medical expert manually segmented subjects: (Lateral Ventricle Volume = 98%, Total CSF Volume = 98%, Intra-Cranial Volume = 99%, Midline Shift Index = 95%). Laboratory testing of CT CoPilot™ segmentation reliability demonstrates equivalent test-retest performance as expert manually segmented subjects. The accuracy testing of CT CoPilot™ registration and segmentation performance was similar to the predicate device (NeuroQuant) without additional safety risk.

CT CoPilot™ consists of proprietary software developed by ZepMed, Inc. installed on an offthe-shelf personal computer. The output of CT CoPilot™ is intended for Picture Archive and Communications System (PACS) display systems. PACS display systems must contain sufficient functionality to display color images (either MR or Secondary Capture), and to display report text and graphics output either in DICOM Structured Report or display of Adobe PDF or JPEG files.

The accuracy of the automatic cross-sectional registration and segmentation in CT CoPilot™ is affected by the subject's deviation from the features embedded in the reference neuroanatomic Atlas. In general, the accuracy of the CT CoPilot™ system software may decrease when the subject's head includes pathologic features not present in the Atlas, lacks features present in the Atlas, or is structurally different than that defined in the preexisting neuroanatomic Atlas. The accuracy of the CT CoPilot™ single image and serial image analysis program may also be degraded by patient motion or by artifacts which are introduced into the patient scanning process, or if patient positioning severely deviates from expectations. Alignment of subject to Atlas space is an important software step in the overall process of registration and segmentation. CT CoPilot™ uses the mechanism for determination of alignment accuracy instantiated in the predicate device (NeuroQuant) for establishing the relationship between image quality and overall alignment accuracy. Specifically, a "measurement index" is determined for each study based on the deviation of each image from the normalized anatomic index embedded in the program. The measurement index is determined by the deviation of the image volumes from normal atlas space. The automatically determined "measurement index" is used to define the limit of anatomic alignment variance which has been determined through laboratory testing of CT CoPilot™ to be accepted by the program. As a safety feature, the CT CoPilot™ software calculates and reports a measurement index number which reflects the adequacy of overall single image and serial image alignment. If the measurement index exceeds specified limits, an error report is generated to inform the user and further processing is terminated.

Factors that may degrade the technical quality and accuracy of CT CoPilot™ registration and segmentation results include:

• a) Patient's motion during scan.
• b) Artifacts affecting overall image quality.
• c) Reconstruction artifacts.
• d) Pathological and or anatomical deviations from the Atlas.
• e) Large initial alignment deviations between the patient and the Atlas.

Here's a breakdown of the acceptance criteria and study information for the CT CoPilot™ device, based on the provided document:

Acceptance Criteria and Device Performance

Note: The document explicitly states acceptance criteria for registration accuracy (less than 15 degrees for the angles). For segmentation accuracy, the acceptance criteria are not numerically defined, but the high correlation coefficients (95-99%) against expert manual segmentation implicitly indicate that these values met the internal performance requirements for substantial equivalence. The reliability criterion is directly stated as "equivalent."

Study Information

2. Sample Size and Data Provenance:
* Test Set Sample Size:
* Registration Accuracy: 100 randomly acquired CT head scans.
* Segmentation Accuracy/Reliability: 179 scans from 34 patients.
* Data Provenance: Retrospective (implied by "randomly acquired CT head scans" and "scans from 34 patients"). The country of origin is not specified. The scans included both normal patients and those with abnormal pathologies, and for segmentation, patients with ventriculostomy catheters.

3. Number of Experts and Qualifications:
* Number of Experts: The document refers to "medical expert manually segmented subjects" for establishing ground truth, implying multiple experts. The exact number is not stated.
* Qualifications of Experts: Not explicitly stated beyond "medical expert."

4. Adjudication Method:
* The document does not explicitly describe an adjudication method like 2+1 or 3+1. It states the correlation coefficients are "when compared against the same group of medical expert manually segmented subjects," suggesting a single set of expert segmentations used as ground truth for comparison.

5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study:
* No, an MRMC comparative effectiveness study that measures the effect size of how much human readers improve with AI vs. without AI assistance was not reported in this document. The study focused on the standalone performance of the device against expert manual segmentation and retest reliability.

6. Standalone Performance:
* Yes, a standalone performance study (algorithm only without human-in-the-loop performance) was conducted. The reported registration accuracy, segmentation accuracy (correlation coefficients), and segmentation reliability directly reflect the device's performance in isolation.

7. Type of Ground Truth Used:
* Expert Consensus/Manual Segmentation: For segmentation accuracy, the ground truth was established by "medical expert manually segmented subjects."
* For registration accuracy, the ground truth appears to be based on an internal reference (e.g., a "normalized anatomic index embedded in the program" or an "Atlas" as mentioned later in the document regarding alignment comparison).

8. Sample Size for the Training Set:
* The document does not specify the sample size used for the training set. It refers to a "reference neuroanatomic Atlas" but does not give details about its derivation or the number of cases used to build it.

9. How the Ground Truth for the Training Set Was Established:
* The document does not explicitly state how the ground truth for the training set was established. It mentions that the accuracy of CT CoPilot™ is affected by the subject's deviation from the "features embedded in the reference neuroanatomic Atlas." This implies the Atlas itself serves as a form of ground truth for some aspects of the algorithm's design, but the origin and ground truth establishment for this Atlas are not detailed. It also mentions "preexisting neuroanatomic Atlas," suggesting it might be an external resource.

Ask a specific question about this device