Using Artificial Intelligence to Help Doctors Identify Different Skin Conditions and Improve Patient Care

A Multi-Reader Multi-Case Study for Evaluating the Impact of Legit.Health Plus Device on the Healthcare Practitioners' Assessment of Skin Lesions

This study aims to determine if an artificial intelligence (AI) medical device can help doctors more accurately identify a wide variety of skin conditions and improve the efficiency of patient consultations. While many patients visit primary care for skin issues, general doctors may sometimes have different opinions from specialists, which can lead to delays in getting the right treatment.

The researchers hypothesized that using the AI tool would increase the true diagnostic accuracy of healthcare professionals for multiple skin conditions. To test this, 16 doctors (including 10 general practitioners and 6 dermatologists) evaluated 29 different medical images.

For each case, the doctors followed a structured process:

• Initial Assessment: Doctors first gave a diagnosis based only on the patient's image and medical history.
• AI Support: Doctors were then shown the AI's top five suggested diagnoses and confidence levels to see if they wished to adjust their final decision.
• Clinical Utility: Doctors also indicated if the patient required a specialist referral and if the case could be handled through a remote (online) consultation.

The primary question the study tried to answer was whether AI support could significantly improve correct diagnoses across 13 different types of skin pathologies-ranging from common rashes to skin cancer-while also making the consultation process faster and more effective for both doctors and patients.

This detailed description outlines the clinical methodology, technical framework, and data integrity protocols utilized in the investigation of the Legit Health Plus medical device for skin pathologies in primary care and dermatology.

Study Design and Technical Methodology The research was conducted as a prospective observational and cross-sectional self-controlled study. It utilized a Multi-Reader Multi-Case (MRMC) framework to evaluate the impact of Computer-Aided Diagnosis (CAD) on clinician performance.

• Self-Controlled Framework: Each healthcare professional (HCP) served as their own comparator, providing diagnoses first without the use of the device and subsequently with the support of the device on the same set of images.
• Evaluation Workflow: Participants accessed a secure web-based platform to review 29 clinical cases. For each case, doctors provided an initial diagnosis based on an image and medical history, followed by a final diagnosis after reviewing the AI's top 5 suggested International Classification of Diseases (ICD) categories and confidence levels.
• Clinical Utility Assessment: The study included a specific questionnaire to evaluate the utility of the data, consultation time reductions, and confidence in making remote clinical decisions.
• Pathology Diversity: The dataset included 13 distinct skin conditions, ranging from common ailments like Acne and Dermatitis to malignant conditions such as Melanoma and Basal Cell Carcinoma.

Quality Assurance and Data Management

To ensure the scientific integrity of the investigation, the following quality and monitoring protocols were implemented:

• Data Validation and Checks: A validation process was carried out by running computer filters based on validation rules to automatically identify missing values or inconsistencies. Manual editing and exploratory statistical techniques were also used to complement error detection.
• Monitoring Plan: The investigation was monitored by a designated clinical monitor, independent of the investigational site, to ensure the protection of subject rights and data accuracy. Monitoring included remote video or telephone meetings every 3 months.
• Bias Minimization: HCPs were randomly selected to ensure that outcomes were not influenced by pre-existing participant characteristics. Standardized procedures for conducting the study and measuring outcomes were strictly followed to reduce variability.

Ethical and Confidentiality Framework The study adhered to ISO 14155:2021, the Declaration of Helsinki, and the General Data Protection Regulation (GDPR).

• Anonymization: All clinical images were completely anonymized and sourced from public dermatological atlases, containing no information that would allow the identification of patients.
• Data Security: All data entries were timestamped, encrypted using industry-standard security protocols, and stored in a secure central database.
• Data Retention: Upon completion of the study and drafting of the final report, all information stored in the device platform is scheduled to be permanently deleted.

Inclusion Criteria:

• Board-certified primary care practitioners and dermatologists, regardless of their professional experience.
• High-quality images of patients with different skin conditions.

Exclusion Criteria:

• Low-quality images of patients which can not be properly analyzed.