Regulations last checked for updates: Nov 22, 2026

Title 21 - Food and Drugs last revised: Nov 12, 2026
§ 864.3900 - Digital cervical cytology slide imaging system with artificial intelligence algorithm.

(a) Identification. The digital cervical cytology slide imaging system with artificial intelligence (AI) algorithm is a prescription in vitro diagnostic device consisting of an automated digital slide creation, viewing, and management system intended to aid in the review of digital images of slides prepared from Pap test specimens and conventional Pap smears by selecting and presenting areas of interest to facilitate interpretation by the reader. The AI software algorithm uses machine learning techniques to provide information to the user about the presence, location, or characteristics of areas of the image with clinical implications. Information from this device is intended to assist the user in rendering a cytology diagnosis.

(b) Classification. Class II (special controls). The special controls for this device are:

(1) The labeling required under § 809.10(b) of this chapter must include the following:

(i) The intended use must include the specimen type (e.g., Pap test or conventional Pap smear), information on the device input (e.g., Pap test glass slides) and output(s) (e.g., whole slide image (WSI) of cytology glass slides or cell spot), information on the AI algorithm output(s) (i.e., format of the information provided by the device to the user, such as objects of interest (OOI) found on the WSI, which is presented in a gallery for the user to review), and intended users;

(ii) A detailed description of the intended patient population;

(iii) Detailed instructions for use, including the intended image review and interpreting protocol for use of the device, and how to resolve device-related issues (e.g., cybersecurity or device malfunction issues);

(iv) A detailed description of the device, including the following:

(A) Necessary input/output devices (e.g., scanner, imager, management server, review station, viewing software, software algorithm);

(B) Necessary infrastructure, platform, and/or software application services, and configuration(s) (e.g., storage, network architecture, cloud services);

(C) Compatible scanning hardware and scanning protocols;

(D) Interaction with input/output devices, and necessary third-party software;

(E) Recommended training for intended users for safe use of the device;

(v) A detailed summary of the performance testing required by paragraphs (b)(2)(i), (b)(2)(ii), (b)(2)(iv), and (b)(2)(v)(A) of this section, including test methods, dataset characteristics, results, and a summary of sub-analyses on case distributions stratified by relevant confounders, i.e., three-dimensional (3D) characteristics of cells and cell clusters in cytology samples, patient demographics, medical history, user experience, sites, and scanning equipment as applicable;

(vi) Intended user screening time data to provide information about daily slide screening workload limits, as determined by the performance testing in paragraph (b)(2)(v)(B); and

(vii) Limiting statements, including the following:

(A) A description of situations in which the device may fail or may not operate at its expected performance level (e.g., poor image quality or for certain subpopulations), including any limitations due to the dataset used to train, test, and tune the algorithm during device development;

(B) That review of the entire gallery (i.e., OOIs with clinical implications) should be performed by the user prior to rendering a diagnosis;

(C) That users should use the device in conjunction with the standard of care, an assessment of patient history and other risk factors, and professional screening and management guidelines to guide patient care; and

(D) That users should employ appropriate procedures and safeguards (e.g., daily quality control measures, deferral to glass slide review, cybersecurity), including when the device is used for remote review of images, to assure the validity of the interpretation of images obtained using this device.

(2) Design verification and validation must include:

(i) A detailed description and the technical performance assessment of each component in the device, including the following, as appropriate:

(A) Slide feeder;

(B) Light source;

(C) Imaging optics;

(D) Z-focal plane acquisition, including a detailed description of methods used to ensure that the complete depth of the cellular material is captured (e.g., Z-plane merging, Z-stacking), when applicable;

(E) Mechanical scanner movement;

(F) Digital imaging sensor;

(G) Image processing software;

(H) Image composition techniques;

(I) Image file formats;

(J) Image review manipulation software;

(K) Computer environment; and

(L) Display system.

(ii) Detailed technical performance assessment at the system level, including for the following, as appropriate:

(A) Coverage of Z-depth to represent the 3D aspects of cells and cell clusters in cytology samples;

(B) Color reproducibility;

(C) Spatial resolution;

(D) Focusing accuracy;

(E) Whole slide cell spot or specimen material coverage;

(F) Stitching error; and

(G) Turnaround time.

(iii) A detailed description of the device software, including its algorithm and its development, that includes a description of any datasets used to train, tune, or test the software algorithm. This detailed description of the device software must include:

(A) The study protocols (e.g., OOI reproducibility study, imaging reproducibility study, cell count validation study) and results used to assess the device output(s) (e.g., image overlays, image heatmaps, image location, total number of cells);

(B) The training dataset, including cases representing different pre-analytical variables representative of the conditions likely to be encountered when used as intended (e.g., fixation type and time, cytology slide processors, challenging diagnostic cases, multiple sites, patient demographics);

(C) An appropriate number of digital cytology images in an independent validation dataset to demonstrate device accuracy in detecting OOI on scanned cytology images, which must include subsets clinically relevant to the intended use of the device; and

(D) A system level architecture diagram with a matrix to depict the communication endpoints, communication protocols, and security protections for the device and its supportive systems, including any products or services that are included in the communication pathway.

(iv) Detailed documentation of studies demonstrating acceptable, as determined by FDA, analytical device performance using samples that are representative of the entire spectrum of challenging cases likely to be encountered when the device is used as intended. For each analytical study, relevant details must be documented (e.g., the origin of the study slides and images, reader/annotator qualifications, method of annotation, location of the study site(s), challenging diagnoses). The analytical studies must include:

(A) Testing to assess device outputs (e.g., OOI reproducibility study, cell count validation study); and

(B) Precision and reproducibility study(ies) that demonstrate device performance when used with multiple input devices (e.g., digital cytology scanners) to assess total variability across images, operators, within-scanners, between-scanners and between-sites, using a comprehensive set of representative clinical specimens with defined, clinically relevant features and challenging characteristics likely to be encountered when the device is used as intended. Precision and reproducibility must be assessed by agreement between replicates.

(v) Data demonstrating acceptable, as determined by FDA, clinical validation by conducting studies with clinical specimens to demonstrate that screening and diagnosing digital images of cervical cytology slides with the assistance of the device's software algorithms is non-inferior to using a comparator method that FDA has determined to be appropriate (e.g., optical microscope). For each clinical study, relevant details must be documented (e.g., the origin of the study slides and images, reader/annotator qualifications, method of annotation, location of the study site(s) (on-site/remote), challenging diagnoses). The studies must include:

(A) A study demonstrating the performance by the intended users with and without the AI algorithm (i.e., unassisted (manual microscope read) and device-assisted reading of scanned digital images of cervical cytology slides). The study dataset must contain sufficient numbers of cytological slides for each diagnostic category and processed by representative processing instruments such that the performance estimates and confidence intervals for these individual subsets can be appropriately characterized. The performance assessment must be based on appropriate diagnostic accuracy measures (e.g., sensitivity, specificity, predictive value, diagnostic likelihood ratio); and

(B) Intended user screening time data obtained from the clinical study, to characterize and set recommended daily maximum workload screening limits for intended users. The study dataset must contain sufficient cases from relevant cohorts that are representative of the scope of patients likely to be encountered given the intended use of the device (e.g., subsets defined by clinically relevant confounders, challenging diagnoses, subsets with potential biopsy appearance modifiers, concomitant diseases, subsets defined by image scanning characteristics). The reader population must be comprised of the intended user population in terms of clinical training, certification, and years of experience.

(vi) Human factors study data to evaluate the usability and user experience of the device with representative users.

[91 FR 69688, Nov. 5, 2026]
authority: 21 U.S.C. 351,360,360c,360e,360j,360
cite as: 21 CFR 864.3900