Computer-aided detection/diagnosis (CAD) is generally defined as systems that assist doctors to interpret medical images and make diagnostic decisions. Advancing along with generation by generation, CAD products have been improving their performance to meet the clinical needs. Notably, the accomplishments of CAD are highly expected since artificial intelligence (AI) technology with high image processing ability has been used in system development.
It is necessary to prove efficacy before a new CAD product can be applied to clinical sites. A good result of the CAD product only by itself may not be enough. (For example, good accuracy in detecting lesions by using the image data is also used in the development.) It may be necessary to demonstrate specifically to acquire the authorities’ approval. The essential question in approval qualifications is:
What kind of benefit can the new CAD system bring to the real clinical world?
If you are considering entering the Japanese market, Micron may be your best choice for a partner to acquire approval. We provide extensive services from the beginning of market positioning and clinical trial design based on the consultation with PMDA, the conduction of clinical performance study (reader study), and to the submission of the final documents for approval.
An Example of Micron’s Workflow: Assisting CAD Approval
Consultation with PMDA
- About market positioning: classification of medical devices (Class I to IV), comparison with the predicate devices, Notification or Approval.
- About the performance test and clinical trial design, e.g. diseases/conditions of interest, endpoints, sample size, statistical analysis, reference standard, etc.
Development of the Study Protocol
- To develop documents including study protocol for the standalone performance assessment, clinical performance assessment and other manuals.
Development of the Reference Standard*
-
-
- To define the reference standard, e.g. assessment result of radiological images by the clinical truthers, pathological test result.
- To create the reference standard data / label.
*Reference Standard: Also known as “gold standard” or “ground truth”.
-
Standalone Performance Study
- To test the detection accuracy by CAD alone by comparing CAD’s output and the reference data.
- To specify characteristics of CAD output, e.g. imaging findings related to false positive or false negative.
- To conduct subgroup analysis based on important covariates, e.g. lesion size.
Pilot Study (Optional)
- To test the clinical performance study procedures, especially in the case of lacking previous supporting evidence.
- The pilot study result helps to estimate the sample size of study population and necessary reader numbers.
Clinical Performance Study / Reader Study
- To use the design of multiple reader multiple case (MRMC) study, comparing the performance results of readers’ interpretation by using CAD (with CAD) and not using CAD (without CAD).
- To arrange the reading sequences, e.g. sequential design, cross-over design.
- To conduct the readers’ training on how to use the CAD while reading and how to understand CAD’s output.
Statistical analysis of the Study Result
- To conduct the receiver operating characteristic (ROC) curve analysis, e.g. ROC, LROC, FROC, AFROC, etc.
- To demonstrate by providing the statistical analysis result of “with CAD” superior to “without CAD”.
Approval Document Submission
- To develop the clinical study report.
- To prepare other necessary documents and data for approval.
Micron’s Work Experience of Assisting CAD Approval
| Disease Area | Work Description |
|---|---|
| Oncology | ✓ Consultation before approval submission |
| Orthopedics | ✓ Assisting in development of the protocol and other documents |
| Oncology | ✓ Consultation with PMDA ✓ Assisting in development of the protocol and other documents ✓ Development of the Reference Standard |
| Oncology | ✓ Development of the Reference Standard |
| Oncology | ✓ Consultation with PMDA ✓ Assisting in development of the protocol and other documents ✓ Development of the Reference Standard ✓ Standalone Performance Study ✓ Preliminary Study and Clinical Performance Study ✓ Statistical analysis ✓ Development of the clinical study report |
| Oncology | ✓ Consultation with PMDA ✓ Development of the Reference Standard ✓ Standalone Performance Study ✓ Preliminary Study and Clinical Performance Study ✓ Statistical analysis ✓ Development of the clinical study report |
| Oncology | ✓ Survey of the substantial equivalence with the predicate devices |
| Oncology | ✓ Survey of the substantial equivalence with the predicate devices |
| Dementia | ✓ ICCC (In Country Clinical Care-taker) |
| Dementia | ✓ Consultation with PMDA ✓ Assisting in development of the protocol and other documents |
As a leading imaging CRO in Japan, Micron has been supporting multiple imaging clinical trials. As a clinical research organization for imaging clinical trials, we always propose the optimal design of the clinical trials as a reliable partner of the Sponsor. Please contact us if you’re interested in the application and registration of regulatory approval (e.g. Shonin/Ninsho) for “CAD with AI” and intelligence decision support systems for diagnosis based on medical imaging.
Recommendations
White Papers