(107 days)
Spectral Bone Marrow is an automated image processing software application, utilizing deep learning technology for bone segmentation, to facilitate optimized visualization of bone marrow in spectral body and extremity CT images. Spectral Bone Marrow's output can be used during the review of traumatic and non-traumatic bone pathologies.
Spectral Bone Marrow is a deep-learning based software analysis package designed for the visualization of bone marrow based on GE HealthCare's spectral CT acquisitions data. Spectral Bone Marrow assists clinicians by providing an automatically generated fused material density image of the segmented bone region over a base monochromatic image optimized for the visualization of bone marrow during the review of traumatic or non-traumatic bone pathologies. The software creates a fully automated post processing workflow for creating these images and improving reader efficiency.
The Spectral Bone Marrow application involves generating a bone mask with a deep learning bone segmentation algorithm and uses this segmented region to define bone regions of water minus hydroxyapatite (Water(HAP)) material density images, which are subsequently colored. The application outputs the colored Water(HAP) material density images overlayed on monochromatic CT images or Virtual Unenhanced (VUE) images.
Additionally, the Spectral Bone Marrow application has the optional ability to automatically set the window width and window level of the color overlay images to provide for optimal visualization of bone marrow. The software provides multiplanar export of the fused images. Spectral Bone Marrow is hosted on GE's Edison Health Link (EHL) computational platform.
Here's a breakdown of the acceptance criteria and study details for the Spectral Bone Marrow device, based on the provided FDA 510(k) summary:
Acceptance Criteria and Device Performance
The provided document does not explicitly present a table of quantitative acceptance criteria for device performance. However, it states that the engineering bench testing showed that the algorithm is capable of accurately segmenting bones and is safe and effective for its intended use. Furthermore, the clinical assessment validated that the Spectral Bone Marrow software provides additional diagnostic value for the evaluation of bone marrow and increased overall reader efficiency.
This implies that the acceptance criteria revolved around demonstrating accurate bone segmentation, diagnostic value, and improved reader efficiency. The specific metrics and thresholds for "accuracy," "additional diagnostic value," and "increased efficiency" are not quantified in this summary.
Inferred Acceptance Criteria:
| Criterion | Reported Device Performance |
|---|---|
| Bone Segmentation Accuracy | The engineering bench testing demonstrated that the algorithm is "capable of accurately segmenting bones." |
| Diagnostic Value | The clinical assessment "validated that the Spectral Bone Marrow software provides additional diagnostic value for the evaluation of bone marrow." The radiologists provided an assessment of "image quality related to diagnostic use according to a Likert scale." |
| Reader Efficiency Enhancement | The clinical assessment "validated...increased overall reader efficiency." Readers were asked to "rate their efficiency when using the algorithm compared to using without." |
| Safety and Effectiveness (Overall) | The engineering bench testing showed the algorithm is "safe and effective for its intended use." The submission concludes that the device is "substantially equivalent and hence as safe and as effective as the legally marketed predicate device." "GE's quality system's design, verification, and risk management processes did not identify any unexpected results, or adverse effects stemming from the changes to the predicate." The substantial equivalence is based on the software documentation for a "Moderate" level of concern. |
Study Details
1. Sample Size Used for the Test Set and Data Provenance
- Test Set Sample Size: 146 retrospective Spectral CT series.
- Data Provenance: Retrospective, with the country of origin not explicitly stated, but the mention of "US board certified radiologists" suggests the data could be from the US or at least interpreted by US experts.
2. Number of Experts Used to Establish Ground Truth for the Test Set and Qualifications
- Number of Experts: Three.
- Qualifications: US board certified radiologists with expertise in both the evaluation of bone marrow and dual energy imaging review.
3. Adjudication Method for the Test Set
The document does not explicitly state a specific adjudication method like "2+1" or "3+1." It mentions that the ground truth for the 146 retrospective Spectral CT series was generated by three US board certified radiologists. This implies an expert consensus approach, but the specific decision-making process (e.g., majority vote, agreement by all three) is not detailed.
4. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- Was an MRMC study done? Yes, a form of clinical assessment involving multiple readers was conducted. The clinical testing involved "three board certified radiologists" assessing a "representative set of clinical sample images" processed by the software. They evaluated "image quality related to diagnostic use according to a Likert scale" and rated "their efficiency when using the algorithm compared to using without." This setup resembles an MRMC study, focusing on reader performance with and without the AI.
- Effect size of human readers improve with AI vs without AI assistance: The document states that the assessment validated that the Spectral Bone Marrow software provides additional diagnostic value for the evaluation of bone marrow and increased overall reader efficiency. However, it does not provide quantitative effect sizes (e.g., specific percentage improvement in diagnostic accuracy or time saved). It merely confirms an improvement was observed and validated.
5. Standalone (Algorithm Only Without Human-in-the-Loop Performance) Study
Yes, a standalone study (referred to as "engineering bench testing") was done to evaluate the bone segmentation algorithm itself. This testing used a database of 146 retrospective Spectral CT series. The "result of the engineering bench testing showed that the algorithm is capable of accurately segment bones and is safe and effective for its intended use."
6. Type of Ground Truth Used
- Expert Consensus: The ground truth for the engineering bench testing (bone segmentation evaluation) was "generated by three US board certified radiologists."
- Clinical Assessment by Experts: For the clinical evaluation of diagnostic value and reader efficiency, the "assessment used retrospectively collected clinical cases" and each image was "read by each board certified radiologist who provided an assessment of image quality related to diagnostic use according to a Likert scale." This also relies on expert interpretation and assessment as the ground truth for these subjective measures.
7. Sample Size for the Training Set
The document does not specify the sample size for the training set used to develop the deep learning bone segmentation algorithm. It only refers to the training of a "deep learning bone segmentation algorithm."
8. 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 only mentions that the bone segmentation algorithm uses deep learning technology. Typically, ground truth for training deep learning models involves manual annotation by experts, but this detail is not provided in the summary.
§ 892.1750 Computed tomography x-ray system.
(a)
Identification. A computed tomography x-ray system is a diagnostic x-ray system intended to produce cross-sectional images of the body by computer reconstruction of x-ray transmission data from the same axial plane taken at different angles. This generic type of device may include signal analysis and display equipment, patient and equipment supports, component parts, and accessories.(b)
Classification. Class II.