Search Results

EFAI HCAPSeg is a software device intended to assist trained radiation oncology professionals, including, but not limited to, radiation oncologists, medical physicists, and dosimetrists, during their clinical workflows of radiation therapy treatment planning by providing initial contours of organs at risk on non-contrast CT images. EFAI HCAPSeg is intended to be used on adult patients only.

The contours are generated by deep-learning algorithms and then transferred to radiation therapy treatment planning systems. EFAI HCAPSeg must be used in conjunction with a DICOM-compliant treatment planning system to review and edit results generated. EFAI HCAPSeg is not intended to be used for decision making or to detect lesions.

EFAI HCAPSeg is an adjunct tool and is not intended to replace a clinician's judgment and manual contouring of the normal organs on CT. Clinicians must not use the software generated output alone without review as the primary interpretation.

EFAI RTSuite CT HCAP-Segmentation System, herein referred to as EFAI HCAPSeg, is a standalone software that is designed to be used by trained radiation oncology professionals to automatically delineate organs-at-risk (OARs) on CT images. This auto-contouring of OARs is intended to facilitate radiation therapy workflows.

The device receives CT images in DICOM format as input and automatically generates the contours of OARs, which are stored in DICOM format and in RTSTRUCT modality. The device does not offer a user interface and must be used in conjunction with a DICOM-compliant treatment planning system to review and edit results. Once data is routed to EFAI HCAPSeg, the data will be processed and no user interaction is required, nor provided.

The deployment environment is recommended to be in a local network with an existing hospital-grade IT system in place. EFAI HCAPSeg should be installed on a specialized server supporting deep learning processing. The configurations are only being operated by the manufacturer:

• Local network setting of input and output destinations;
• Presentation of labels and their color; ●
• Processed image management and output (RTSTRUCT) file management. ●

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

Acceptance Criteria and Device Performance

Study Details

1. Sample Size and Data Provenance (Test Set):

   • Sample Size: 157 non-contrast CT cases.
   • Data Provenance: Consecutively collected from the United States (U.S.). All data was acquired independently from product development training and internal testing. This indicates a prospective-like collection for validation.
   • Demographics:
     • 30.57% females, 57.96% males, 11.46% gender not reported.
     • Mean age: 61.69 years (SD 11.90 years).
     • CT Manufacturer: GE (43.31%), Philips (36.30%), Siemens (9.55%), Toshiba (2.55%), PNS (0.64%), not reported (7.64%).
     • Location, Race, and Ethnic distribution: Unavailable.
     • Cancer types: Head-and-neck, pancreas, colorectal, breast, bladder, prostate, stomach, gynecologic, sarcoma, and metastatic tumors from multiple origins.

2. Number of Experts and Qualifications (Ground Truth for Test Set):

   • Number of Experts: Three (3) board-certified radiation oncologists.
   • Qualifications: "Board-certified radiation oncologists." (Specific years of experience are not mentioned, but board certification implies a high level of expertise). Adjudication method is described next.

3. Adjudication Method (Test Set):

   • The OAR contouring for the ground truth was generated by "three board-certified radiation oncologists as the ground truth (GT)." The text implicitly suggests a consensus or independent review that established the GT, but it doesn't specify a formal adjudication method like "2+1" or "3+1". It simply states that their combined work constituted the GT.

4. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study:

   • No, a MRMC comparative effectiveness study focusing on human readers improving with AI vs. without AI assistance was not explicitly described.
   • The study performed a standalone performance test comparing the algorithm's performance (EFAI HCAPSeg) against a "comparison device" (another algorithm/device, not human readers). The "comparison device" results are used as a benchmark for EFAI HCAPSeg's performance.

5. Standalone Performance Study (Algorithm Only):

   • Yes, a standalone performance test was done.
   • Method: The EFAI HCAPSeg's OAR contouring capabilities were compared against a "comparison device."
   • Results (EFAI HCAPSeg vs. Comparison Device, Mean DSC):
     • Head & Neck: 0.80 (EFAI HCAPSeg) vs. 0.75 (Comparison Device)
     • Chest: 0.90 (EFAI HCAPSeg) vs. 0.84 (Comparison Device)
     • Abdomen & Pelvis: 0.90 (EFAI HCAPSeg) vs. 0.82 (Comparison Device)
     • Unique OARs of EFAI HCAPSeg: 0.82 (compared against a pre-specified value, not the comparison device)
   • Results (EFAI HCAPSeg vs. Comparison Device, Median 95% HD mm):
     • Head & Neck: 2.17 (EFAI HCAPSeg) vs. 3.09 (Comparison Device)
     • Chest: 2.23 (EFAI HCAPSeg) vs. 3.87 (Comparison Device)
     • Abdomen & Pelvis: 2.28 (EFAI HCAPSeg) vs. 3.90 (Comparison Device)
     • Unique OARs of EFAI HCAPSeg: 2.24 (compared against a pre-specified value)

6. Type of Ground Truth Used (Test Set):

   • Expert Consensus: The ground truth was established by "three board-certified radiation oncologists" who manually contoured each Organ-at-Risk (OAR).

7. Sample Size for Training Set:

   • Total Cases: 1,410 adult cases.
   • Number of Images: Varies significantly by OAR, ranging from hundreds to over 100,000 images per OAR (e.g., SpinalCord had 139,337 images). (Refer to Table C for detailed counts per OAR).

8. How the Ground Truth for the Training Set Was Established:

   • The text states, "The process of ground-truthing, involving the manual contouring of each OAR, was undertaken by three board-certified radiation oncologists." It further adds, "The data collection and ground truth protocol was done following the identical procedures as those of the predicate device." While not explicitly stated for the training set ground truth, it is highly implied that the same process (manual contouring by three board-certified radiation oncologists) was used for both training and testing datasets for consistency, especially given the statement about identical procedures to the predicate device. The demographic information in Table B covers both training and testing datasets, reinforcing this.

Ask a specific question about this device