Computer-Assisted Detection Devices: Definition and Features
The new guidance issued by the Agency is intended to assist medical device manufacturers and other parties in assessing compliance of computer-assisted detection (CADe) devices. Devices of this kind can analyze information obtained from the patient, like radiological data, and recognize patterns, allowing them to identify abnormalities on the radiology image. Due to their complexity, such devices should be used by healthcare professionals.
The FDA mentions that the document is based on the feedback and comments obtained in the course of public consultations regarding the related issues performed during the previous several years. The Agency also reminds that in accordance with the special FDA Order certain medical image analyzers were assigned to Class II and thus should be subject to premarket notification requirements while the present guidance also provides additional information that could be helpful for medical devices manufacturers maintaining compliance with the aforementioned special controls requirements.
Radiological data provided by the devices covered by the guidance include images obtained from the patient’s examination using ultrasound, radiography, magnetic resonance imaging (MRI), positron emission / computed tomography. The scope of the guidance covers all CADe devices (even ones being a part of a combined system). In the terms of the nomenclature and product codes, the guidance is dedicated to devices covered by 21 CFR 892.2050 “Picture archiving and communications systems” and 21 CFR 892.2070 “Medical image analyzer” (NWE, OEB, MYN). At the same time, the authority mentions that the new medical device codes may be introduced if found reasonably necessary to classify new devices.
As usual, each particular CADe device contains one scheme intended to identify one particular type of abnormality. At the same time, the CADe device can also contain several detection schemes simultaneously. In general, the guidance covers, among others, the following:
- CADe devices intended for the identification of colonic polyps,
- CADe intended for the identification of filling defects using the computed tomography, and
- CADe intended for the identification of brain lesions in the course of magnetic resonance imaging.
Such algorithms could also be amended or modified. In particular, it is possible to change its features, methods used to train it or particular data sets used in the training process, parameters of an algorithm or specific image processing components. At the same time, the FDA explicitly excludes from the scope of the guidance clinical performance assessment studies related to devices used in the course of intra-operative procedures, radiological computer-assisted diagnostic (CADx) software and computer-aided triage and notification software devices except when they are one part to the device that falls within the scope of the document.
CADe Device Regulation
According to the applicable regulations, the manufacturers of Medical Image Analyzers should provide a detailed description of pre-specified performance testing methods. It is also required to provide the data sets utilized when assessing whether the application of the device allows for improvement of the performance of a healthcare professional analyzing radiological data. The information should also include references to the applicable accuracy measurers the assessment is based on. The number of cases in each dataset used for testing should be sufficient for the results to be reliable.
All studies described in the guidance should be enclosed as a part of the 510(k) submission. The Agency also states that in the case of any additional questions arising, an applicant is allowed to contact the FDA in accordance with the procedure set forth for the appropriate request submissions. The Agency also advises the manufacturers to submit their testing protocols to the FDA for preliminary review before conducting both performance assessment and clinical performance assessment, and there is a special form, a Pre-Submission, intended to be used in such cases.
The clinical performance assessment should be performed to confirm that the device meets safety and effectiveness requirements when used for the purpose intended by the manufacturer and its user (healthcare professional) acts strictly in accordance with the instructions supplied with the device and fulfills all the requirements and rules specified therein.
The Agency notifies that there are numerous factors that could potentially impact the actual performance of a device and thus, the assessment of its effectiveness. Due to the way the device should be used, a huge role should be played by a particular specialist (healthcare professional) in operating the device since he directly interprets the information. It is necessary to access various factors, namely:
- Experience of the specialist and his knowledge of abnormalities of the particular type the CADe device is intended to identify,
- Number of marks and their characteristics, including the proximity of abnormality,
- Particular step at which the CADe device is being applied (for example, it could process the data simultaneously with the specialists, i.e. concurrently, or could be used for second reading).
Any clinical performance assessment should be designed in a way that allows avoidance of possible bias that could result in incorrect conclusions regarding compliance with applicable requirements related to safety and effectiveness. The Agency also warns that laboratory testing results could be impacted substantially by unintended deviations since it is quite difficult to create conditions similar to the real clinical environment. To deal with such issues and to improve the accuracy and reliability of the results, the FDA recommends medical device manufacturers to:
- Test devices in real conditions to define and evaluate its real performance (except the cases when the abnormality that the device is intended to identify is too rare so it is difficult to collect the volume of information sufficient to ensure the reliability of testing results),
- Retrospective analysis using the dataset composed with the existing information,
- A stress test – special procedure when the CADe device is subject to testing with the most challenging and complicated cases (e.g. radiological images and data that are quite rare in the course of normal clinical routine).
Another approach used to evaluate the accuracy is Multiple Readers Multiple Cases (MRMC) providing that the same information should be simultaneously processed by several specialists (healthcare professionals) in different conditions. For example, some of the specialists use CADe devices to process the information when some don’t. To ensure the highest possible accuracy and reliability of results, all specialists should process all the data and examine all the cases. However, the latter requirement is not obligatory under certain conditions and the results of the simplified study are also acceptable.
CADe Application Components and Information to be Provided
The document also contains a list of documents to be provided by the applicant in the course of filing 510(k) application submission. First of all, the Agency emphasizes the importance of providing clinical study protocols, that should be provided for the review in advance. The applicant shall also provide:
- A detailed description of the design of the study performed,
- Details regarding the data collection process including the model of the device used,
- Sample of the study protocol that should contain the information on hypothesis tested and initial endpoints, alternative tests, pre-defined success criteria, explanation of the case sample size and the number of specialists engaged in the study, the methodology used to interpret the information together with the methods of randomization and tasks provided to the specialists,
- Qualification of specialists engaged in testing,
- A detailed description of the training process,
- Description of statistical methods used.
How Can RegDesk Help?
RegDesk is a next-generation web-based software for medical device and IVD companies. Our cutting-edge platform uses machine learning to provide regulatory intelligence, application preparation, submission and approvals management globally. Our clients also have access to our network of over 4000 compliance experts worldwide to obtain verification on critical questions. Applications that normally take 6 months to prepare can now be prepared within 6 days using RegDesk Dash(TM). Global expansion has never been this simple.