(88 days)
Not Found
Yes
The device description explicitly states that the device uses artificial intelligence to analyze chest radiographs.
No
The device is described as aiding in the detection of pulmonary nodules, not as treating or preventing a disease or condition. It is a diagnostic aid, not a therapeutic device.
No
The text explicitly states multiple times that "The device is not intended to be used for diagnosis."
Yes
The device description explicitly states "The device is a software device that uses artificial intelligence to analyze chest radiographs." There is no mention of any accompanying hardware components being part of the device itself.
Based on the provided information, this device is not an IVD (In Vitro Diagnostic).
Here's why:
- IVD definition: In vitro diagnostics are tests performed on samples taken from the human body, such as blood, urine, or tissue, to detect diseases, conditions, or infections.
- Device's function: This device analyzes images (chest radiographs) to aid in the detection of pulmonary nodules. It does not analyze biological samples.
- Intended Use: The intended use clearly states it's for aiding in the detection of pulmonary nodules on chest radiographs, not for analyzing biological specimens.
Therefore, this device falls under the category of medical imaging software or a medical device that processes images, rather than an in vitro diagnostic device.
No
The provided input explicitly states "Control Plan Authorized (PCCP) and relevant text: Not Found," indicating no mention of an FDA-reviewed or cleared PCCP for this device.
Intended Use / Indications for Use
The BoneView® system is an entirely software-based computer-assisted detection (CADe) device. It is intended to process 2D projection radiographs (X-ray) of adult appendicular skeleton (i.e., wrists, hands, forearms, elbows, arms, shoulders, feet, ankles, lower legs, knees, femurs, and hips) to identify and mark regions of interest (ROI) that could contain fractures. The device is intended to be used as a concurrent reading aid to assist orthopedic or emergency room physicians, general practitioners, physician assistants, radiologists, and other medical practitioners proficient in interpreting musculoskeletal radiographs, to improve fracture detection. The output of the device is intended to be used in conjunction with full clinical history and other diagnostic test results. The device is not intended as a primary diagnostic device.
Product codes
QBU
Device Description
BoneView® system is an entirely software-based computer-assisted detection (CADe) device. It is intended to process 2D projection radiographs (X-ray) of adult appendicular skeleton (i.e., wrists, hands, forearms, elbows, arms, shoulders, feet, ankles, lower legs, knees, femurs, and hips) to identify and mark regions of interest (ROI) that could contain fractures. The device is intended to be used as a concurrent reading aid to assist orthopedic or emergency room physicians, general practitioners, physician assistants, radiologists, and other medical practitioners proficient in interpreting musculoskeletal radiographs, to improve fracture detection. The output of the device is intended to be used in conjunction with full clinical history and other diagnostic test results. The device is not intended as a primary diagnostic device. The BoneView® CADe system for fracture detection is composed of a software application hosted in a cloud environment accessible via a secure web portal. It receives DICOM images, processes them using artificial intelligence (AI) algorithms, and creates a DICOM Secondary Capture image as output, displaying detected fracture regions along with their probability. This output image can be transmitted back to the facility's PACS system. It is designed to operate seamlessly within existing clinical workflows, providing results directly to healthcare professionals' workstations.
Mentions image processing
Yes
Mentions AI, DNN, or ML
Yes
Input Imaging Modality
2D projection radiographs (X-ray)
Anatomical Site
appendicular skeleton (i.e., wrists, hands, forearms, elbows, arms, shoulders, feet, ankles, lower legs, knees, femurs, and hips)
Indicated Patient Age Range
adult
Intended User / Care Setting
orthopedic or emergency room physicians, general practitioners, physician assistants, radiologists, and other medical practitioners proficient in interpreting musculoskeletal radiographs / Not Found
Description of the training set, sample size, data source, and annotation protocol
The BoneView AI algorithm was developed using a diverse dataset of adult appendicular X-ray images. The training dataset consists of 456,839 images. These images were collected from multiple sources, including existing clinical partners (hospitals in France) under an ethical approval provided by an ethical committee where patients provided informed consent for the use of their anonymized medical data for research. Additional images were collected through agreements with US hospitals that provided de-identified images and public datasets. The images in the training dataset were reviewed and annotated both for the specific body parts represented and the presence or absence of a fracture by highly-trained medical experts (board-certified radiologists or orthopedic surgeons) and an experienced image engineer with clinical knowledge. The annotations provided for each fracture include classification of the type of fracture and a contour of the fracture region. The annotations were conducted through a collaborative process among multiple readers to reach a consensus. These annotations were specifically used for training and optimizing the deep learning models. Quality control on the annotated data was performed for all images and annotations prior to their inclusion in the training dataset.
Description of the test set, sample size, data source, and annotation protocol
A total of 1500 fracture-positive cases and 1500 fracture-negative cases were included in the independent validation dataset. The device was evaluated on 3,000 images, corresponding to 3,000 distinct patients, collected from various sources within France and the US. These images originate from distinct hospitals and are representative of real-world clinical practice because they include a wide distribution of image quality, fracture types, and anatomical sites for fractures. The test set was fully independent from the training dataset. For the validation study, all ground truth labels were first established through an adjudication process with 3-5 board-certified radiologists who independently reviewed each X-ray image. Annotations of fracture locations were made by drawing bounding boxes. In case of disagreement, a centralized adjudication meeting was conducted until a consensus was reached, representing the true underlying disease status, which was used as the ground truth.
Summary of Performance Studies (study type, sample size, AUC, MRMC, standalone performance, key results)
A multi-reader multi-case (MRMC) study was designed to evaluate whether the use of BoneView® CADe will lead to improved fracture detection performance in readers. The study was conducted in a blinded, randomized, and counter-balanced manner. The study included 25 readers, consisting of 5 board-certified radiologists, 5 emergency room physicians, 5 orthopedic surgeons, 5 general practitioners, and 5 physician assistants. Each reader interpreted 300 cases in two reading sessions (unaided and aided by the device). The 300 cases included a diverse set of real-world images (150 fracture-positive and 150 fracture-negative) from the test set of 3,000 images described above. The primary endpoint for this study was the improvement in the Area Under the ROC curve (AUC). A standalone performance study was also conducted to demonstrate the clinical performance of the device without human interpretation compared to the ground truth. The standalone performance was evaluated using the 3,000 images from the independent test set described above. The device standalone performance was evaluated by computing the Area Under the FROC curve, and sensitivity and specificity at the operating point of the device.
Key results from the MRMC study: The results show a statistically significant improvement in the diagnostic accuracy when using BoneView® CADe as an aid. The mean ROC AUC improved from 0.817 (95% CI: 0.787, 0.844) to 0.875 (95% CI: 0.848, 0.900) when readers used BoneView® CADe, resulting in an average AUC improvement of 0.058 (95% CI: 0.042, 0.078, p
§ 880.6250 Non-powdered patient examination glove.
(a)
Identification. A non-powdered patient examination glove is a disposable device intended for medical purposes that is worn on the examiner's hand or finger to prevent contamination between patient and examiner. A non-powdered patient examination glove does not incorporate powder for purposes other than manufacturing. The final finished glove includes only residual powder from manufacturing.(b)
Classification. Class I (general controls). The device, when it is a finger cot, is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 880.9.
0
1
2