Search Results

The qXR-LN (qXR Lung nodule) is computer-aided detection software to identify and mark regions in relation to suspected pulmonary nodules from 6 to 30 mm in size. The device is intended to be used in the incidental adult population. It is designed to aid the physician to review the frontal (AP/PA) chest radiographs of adults acquired on digital radiographic systems as a second reader and be used with any DICOM viewer or PACS . qXR-LN provides adjunctive information only and is not a substitute for the original chest radiographic image.

qXR-LN is a Computer-Aided Detection (CADe) device that is designed to perform CAD processing in frontal (PA or AP view) Chest X-ray images for indication of locations for high nodule probability, which has an effective detection size from 6 mm to 30 mm. The device is intended to be a secondary aid to the qualified intended user to identify incidental pulmonary lung nodules chest x-ray images.

The device utilizes a deep learning algorithm. The qXR-LN was trained on a large and diverse dataset of 2.5million scans from 5 countries across the world. The training dataset was from more than 25 manufacturers.

Chest X-rays are sent to qXR-LN by the means of transmission functions within the user's image storage system (e.g., Picture Archiving and Communication System (PACS)) or other radiological imaging equipment (e.g., X-ray systems) and processed by the qXR-LN to detect and localise lung nodules. Following receipt of chest radiographs, the software device automatically analyses each image to detect and localise lung nodules.

qXR-LN receives chest x-ray images in digital imaging and communications in medicine (DICOM) as input. The qXR-LN device produces DICOM format outputs that enable users to view the presence and location of lung nodules.

This device is intended to aid the intended user in review of chest x-rays and detect and localise lung nodules as a secondary reader. The results are not intended to be used on a standalone basis for clinical decision-making nor is it intended to rule out the target conditions or otherwise preclude clinical assessment of X-ray cases.

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

Acceptance Criteria and Device Performance

1. Table of Acceptance Criteria and Reported Device Performance

The acceptance criteria for qXR-LN are not explicitly stated as distinct numerical targets in the same way performance criteria often are. Instead, the document compares its performance to a predicate device (Samsung Auto Lung Nodule Detection) and demonstrates non-inferiority or improvement. The core principle for acceptance is "substantial equivalence" to the predicate, with performance metrics being a key factor in proving this equivalence.

Based on the provided information, the implicit acceptance criteria are framed around demonstrating performance at least equivalent to the predicate, particularly in terms of improving human reader performance and achieving a competitive standalone sensitivity for nodule detection.

Study Details Proving Acceptance

2. Sample Size Used for the Test Set and Data Provenance

• Test Set Sample Size: The exact number of cases in the test set is not explicitly stated. However, the study "scans were obtained from 8 states (Ohio, New York, South Carolina, Iowa, Wisconsin, Texas, Oklahoma and Maryland) and 40 sites (each state had multiple sites) across the US." The document mentions a "diverse dataset of 2.5 million scans" for training, implying a distinct, smaller, and well-characterized dataset for testing.
• Data Provenance:
  • Country of Origin: United States (8 states, 40 sites).
  • Retrospective or Prospective: Retrospective. The study states "A fully crossed multi-case, multi-reader, retrospectively study design was utilized."

3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications

• Number of Experts: 5 ABR (American Board of Radiology) certified ground truthers.
• Qualifications: ABR certified radiologists. No specific years of experience are mentioned, but ABR certification implies a certain level of expertise.

4. Adjudication Method for the Test Set

The document states, "The standalone study was performed to compare qXR-LN's performance against a ground truth determined by 5 ABR certified ground truthers." It further specifies that "They read the Chest X-rays with the accompanying CT scans and reports and the ground truth was based on the nodules visible on the Chest X-ray." This strongly suggests a consensus-based adjudication method among the 5 experts, likely with a pre-defined process for resolving discrepancies, although the specific "2+1" or "3+1" type is not detailed. The use of accompanying CT scans and reports likely aided in this consensus.

5. If a Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study was done, and its effect size

• Was it done? Yes, a "fully crossed multi-case, multi-reader, retrospectively study design was utilized."
• Effect Size of Human Reader Improvement with AI vs Without AI Assistance:
  • AFROC (Area Under the Free-Response Receiver Operating Characteristic) Improvement: The AFROC of readers was improved by 0.07621 (95% CI: 0.0497 – 0.1026) when aided by qXR-LN compared to unaided, which was statistically significant (P < 1x10-5).
  • Nodule Level Sensitivity Improvement: qXR-LN indicated a 11.96% improvement for nodule level sensitivity when readers were aided.

6. If a Standalone (Algorithm Only Without Human-in-the-Loop Performance) was done

• Was it done? Yes, "A standalone study to assess device performance was conducted."

7. The Type of Ground Truth Used

The ground truth for the test set was established by expert consensus (5 ABR certified radiologists) leveraging multiple modalities:

• Chest X-rays
• Accompanying CT scans
• Reports

The ground truth was specifically "based on the nodules visible on the Chest X-ray," confirmed/localized with CT data. This suggests a blend of expert consensus and validation against a higher fidelity imaging modality (CT), with clinical reports providing further context.

8. The Sample Size for the Training Set

• Training Set Sample Size: "a large and diverse dataset of 2.5 million scans from 5 countries across the world."

9. How the Ground Truth for the Training Set was Established

The document does not explicitly detail how the ground truth for the training set was established. It only mentions the size and diversity of the dataset. Given the scale (2.5 million scans), it's highly probable that a combination of methods was used, potentially including:

• Automated extraction from radiology reports.
• "Weak supervision" techniques where general labels infer detailed annotations.
• A subset of manual expert review.
• Potentially, also leveraging CT scans and reports as done for the test set, but for a smaller, expert-annotated subset used for ground-truthing in training.

However, since this information is not provided in the text, it remains an unknown based on the supplied document.

Ask a specific question about this device