Diagnostic Precision of the AI Tool Dermalyzer to Identify Malignant Melanomas in Subjects Seeking Primary Care for Melanoma-suspected Cutaneous Lesions

Background: Malignant melanoma is a worldwide health issue of concern, of increasing incidence. Early melanoma detection is crucial for survival and prognosis. Dermalyzer is a device intended to be used as a decision support system for assessing cutaneous lesions suspected of being melanomas, based on artificial intelligence (AI), often referred to as "machine learning". The input from the device is not intended to be used as the sole source of information for diagnosis. Intended to be used by medical professionals. The service does not provide any other diagnosis. Study aims/objectives: Primary objective: The primary objective of the investigation is to determine the diagnostic precision of the device; to answer at which level the AI tool Dermalyzer can identify malignant melanomas among cutaneous lesions that are assessed in clinical use due to any degree of malignancy suspicion. Secondary objectives: A) To evaluate usability and applicability in clinical praxis of Dermalyzer by users (medical professionals), and B) To gain an increased knowledge and understanding of how digital tools enhanced co-artificial intelligence can assist physicians with the right support for an earlier diagnosis of malignant melanoma. Exploratory objective: To explore health economic aspects of improved diagnosis support Exploratory endpoints Materials \& methods: The subjects will be included from around 30 primary care centers in Sweden. If the subject's lesion(s) is suspected of melanoma or melanoma cannot be ruled out, the subject is asked to participate in the investigation. The investigator examines the subject's lesion(s) and makes the clinical assessment of the subject lesion(s) based on established clinical decision algorithms (such as "Chaos \& clues", "3- or 7-point checklist", or the ABCDE concept) of whether there is a suspicion of MM, according to the usual clinical routine (also includes very low suspicion of MM but cannot be completely dismissed). The investigator takes dermoscopy images according to standard of care and archives the image(s) according to clinical routine. The investigator decides on action, based on his or her MM suspicion (excision at the primary care center or referral for excision or referral to a dermatologist for further assessment). If the subject has agreed to participate in the investigation, the investigator indicates in the CRF the clinical suspicion level of MM, and decided action. The investigator takes images of the lesion(s) again, this time with a mobile phone, containing the IMD AI SW, connected to a dermatoscope, and follows the on-screen instructions. The image is processed by the AI SW and the results are visible on the screen within seconds. A unique auto generated code number is also presented. The code number is registered on the enrollment log and in the CRF. The investigator records how he considers that the degree of suspicion of MM (higher vs lower) would have been affected by the AI SW result if it had been the governing body for the treatment. When the subject has been fully examined and receives the final tumor diagnosis from the histopathology/PAD results or dermatologist assessment (melanoma/non melanoma), the degree of agreement between the PAD and the outcome of the AI SW decision support is calculated with the Kappa-analysis and the diagnostic accuracy to be able to distinguish melanoma from non-melanoma in the form of sensitivity and specificity as well the positive and predictive value. The corresponding comparison is performed from the examining investigators estimated clinical degree of suspicion, as well as the diagnostic accuracy when both the PAD and the AI decision support are wigheted together (ei in cases where the investigator and the decisions support are in agreement in their assessment). The clinical investigation will collect information from the users, how participating users (investigators at the site) experience the usability of the AI SW decision support and attaching applications, from short surveys including the validated System Usability Scale.