EDDA: Explainable Domain-Specific Artificial Intelligence for Dermatopathological Melanoma Detection — A Diagnostic Accuracy Study

EDDA: Explainable Domain-Specific Artificial Intelligence for Dermatopathological Melanoma Detection — A Diagnostic Accuracy Study - EDDA

Status

Active, not recruiting

Phases

Unknown

Study type

Observational

Source

DRKS

Registry ID

DRKS00034729

Enrollment

1600

Registered

2024-07-25

Start date

2021-04-01

Completion date

Unknown

Last updated

2025-04-07

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

C43 D22

Interventions

Group 1: Patients (1) who presented with at least one melanoma suspicious lesion in one of eight German university hospitals between April 2021 and March 2023 and who met the inclusion criteria of th

Eligibility

Sex/Gender

All

Age

18 Years to No maximum

Inclusion criteria

Inclusion criteria: * presence of clinically melanoma-suspicious skin lesions that were excised following dermoscopic examination

Exclusion criteria

Exclusion criteria: suspicious lesions that * have been previously pre-biopsied * were located near the eye * were located under the fingernails or toenails * have person-identifying features (e.g. tattoos) in their immediate vicinity

Design outcomes

Primary

Measure	Time frame
Diagnostic accuracy of the index test (explaibable AI system EDDA) on the diagnosis of malignant (invasive melanoma and non-invasive melanomas (melanoma in-situ and lentigo maligna)) and benign (nevus) lesions. Malignant and benign lesions are present in the data set at a ratio of 1:1. The diagnostic accuracy is measured by * Accuracy * Sensitivity * Specificity * Area under the receiver-operator curve (AUROC) * Area under the precision-recall curve (AUPRC) With the majority vote of the pathologist panel serving as reference standard.	—

Measure

Time frame

Diagnostic accuracy of the index test (explaibable AI system EDDA) on the diagnosis of malignant (invasive melanoma and non-invasive melanomas (melanoma in-situ and lentigo maligna)) and benign (nevus) lesions. Malignant and benign lesions are present in the data set at a ratio of 1:1. The diagnostic accuracy is measured by * Accuracy * Sensitivity * Specificity * Area under the receiver-operator curve (AUROC) * Area under the precision-recall curve (AUPRC) With the majority vote of the pathologist panel serving as reference standard.

—

Secondary

Measure	Time frame
Diagnostic accuracy of EDDA on the differential diagnosis of invasive melanoma, non-invasive melanoma (NIM; here: melanoma in-situ and lentigo maligna) and nevus. The diagnostic classes are present in a ratio of 3:1:4 (invasive melanoma:NIM:nevus) in the dataset. Due to the number of classes and/or class imbalances, the diagnostic accuracy for the outcomes of the secondary endpoint are measures as follows: * Accuracy * Class-balanced accuracy * F1-score * Sensitivity (using a one-against-all-approach) * Specificity (using a one-against-all-approach) * Area under the receiver-operator curve (AUROC) (calculated per class) * Area under the precision-recall curve (AUPRC) (calculated per class) With the majority vote of the pathologist panel serving as reference standard.	—

Measure

Time frame

Diagnostic accuracy of EDDA on the differential diagnosis of invasive melanoma, non-invasive melanoma (NIM; here: melanoma in-situ and lentigo maligna) and nevus. The diagnostic classes are present in a ratio of 3:1:4 (invasive melanoma:NIM:nevus) in the dataset. Due to the number of classes and/or class imbalances, the diagnostic accuracy for the outcomes of the secondary endpoint are measures as follows: * Accuracy * Class-balanced accuracy * F1-score * Sensitivity (using a one-against-all-approach) * Specificity (using a one-against-all-approach) * Area under the receiver-operator curve (AUROC) (calculated per class) * Area under the precision-recall curve (AUPRC) (calculated per class) With the majority vote of the pathologist panel serving as reference standard.

—

Countries

Germany

Contacts

Public ContactSarah Haggenmüller

DKFZ Heidelberg

sarah.haggenmueller@dkfz-heidelberg.de+49 6221 42 5301

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: Feb 4, 2026