Dermatologist-like explainable AI enhances melanoma diagnosis accuracy: eye-tracking study

Chanda T, Haggenmueller S, Bucher TC, Holland-Letz T, Kittler H, Tschandl P, Heppt M, Berking C, Utikal JS, Schilling B, Buerger C, Navarrete-Dechent C, Goebeler M, Kather JN, Schneider CV, Durani B, Durani H, Jansen M, Wacker J, Wacker J, Kalski M, Klifo D, Kiefer S, Klifo H, Funk T, Lunderstedt J, Buchinger A, Erdogdu U, Weberschock T, Gosmann J, Sachweizer A, Loos S, Fahimi S, Christ F, Dionysia D, Yilmaz K, Ninosu N, Schaarschmidt ML, Baumert J, Sackmann T, Rabe L, Höner M, Zieringer L, Uebel C, Breakell T, Sagonas I, Bosch-Voskens C, Sollfrank L, Ronicke M, Kemenes S, Sambale J, Wagner N, Erdmann M, Ammar AM, Manuelyan K, Salerni G, Rácz E, Saa SR, Hoorens I, Salava A, Lengyel Z, Balcere A, Jocic I, Zafirovik Z, Dragolov M, Hudson S, Cenk H, Tsakiri A, Petrovska L, Neto RRO, Ferhatosmanoğlu A, Morales-Sánchez MA, Bondare-Ansberga V, Afiouni R, Erdil DI, Beyens A, Lluch-Galcerá JJ, Vucemilovic AS, Theofilogiannakou P, Sławińska M, Garzona-Navas L, Hartmann D, Ludwig-Peitsch W, Thamm J, Pföhler C, Hoffmann F, Maul JT, Nguyen VA, Braun SA, Gössinger E, Mühleisen B, Feldmeyer L, Bechara FG, Schuh S, Reimer-Taschenbrecker A, Maul LV, Dimitriou F, Persa OD, Welzel J, Ahlgrimm-Siess V, Booken N, Brinker TJ (2025)

Publication Type: Journal article

Publication year: 2025

Journal

Book Volume: 16

Article Number: 4739

DOI: 10.1038/s41467-025-59532-5

Abstract

Artificial intelligence (AI) systems substantially improve dermatologists’ diagnostic accuracy for melanoma, with explainable AI (XAI) systems further enhancing their confidence and trust in AI-driven decisions. Despite these advancements, there remains a critical need for objective evaluation of how dermatologists engage with both AI and XAI tools. In this study, 76 dermatologists participate in a reader study, diagnosing 16 dermoscopic images of melanomas and nevi using an XAI system that provides detailed, domain-specific explanations, while eye-tracking technology assesses their interactions. Diagnostic performance is compared with that of a standard AI system lacking explanatory features. Here we show that XAI significantly improves dermatologists’ diagnostic balanced accuracy by 2.8 percentage points compared to standard AI. Moreover, diagnostic disagreements with AI/XAI systems and complex lesions are associated with elevated cognitive load, as evidenced by increased ocular fixations. These insights have significant implications for the design of AI/XAI tools for visual tasks in dermatology and the broader development of XAI in medical diagnostics.

Authors with CRIS profile

Involved external institutions

Luzerner Kantonsspital (LUKS) Switzerland (CH) Ruhr-Universität Bochum (RUB) Germany (DE) Universitätsklinikum Mannheim / University Medical Centre Mannheim (Universitätsmedizin Mannheim) Germany (DE) Universitätsklinikum Augsburg Germany (DE) Universitätsspital Zürich (USZ) Switzerland (CH) Al-Azhar University Egypt (EG) Universitätsspital Basel Switzerland (CH) Northwestern University United States (USA) (US) Riga 1st hospital / Rīgas 1. slimnīca Latvia (LV) Military Medical Academy (Serbia) / Војномедицинска академија / Vojnomedicinska akademija (VMA) Serbia (RS) Universitätsklinikum Frankfurt am Main (KGU) Germany (DE) Taksim Eğitim ve Araştırma Hastanesi Turkey (TR) Universitätsklinikum Würzburg Germany (DE) Universitätsklinikum Aachen (UKA) Germany (DE) University Medical Center Groningen (UMCG) / Universitair Medisch Centrum Groningen Netherlands (NL) Helsinki University Central Hospital (HUCH) / Helsingin seudun yliopistollinen keskussairaala (HYKS) Finland (FI) Pamukkale University / Pamukkale Üniversitesi (PAU) Turkey (TR) Deutsches Krebsforschungszentrum (DKFZ) Germany (DE) University of Pécs / Pécsi Tudományegyetem Hungary (HU) Medizinische Universität Wien Austria (AT) Universidad Nacional de Rosario (UNR) Argentina (AR) Saints Cyril and Methodius University of Skopje / Универзитет „Св. Кирил и Методиј“ во Скопје Republic of North Macedonia (MK) University Hospital Split Croatia (HR) Vivantes - Netzwerk für Gesundheit GmbH Germany (DE) Medical University Gdansk / Gdański Uniwersytet Medyczny Poland (PL) München Klinik gGmbH Germany (DE) Hospital Clínica Bíblica Costa Rica (CR) Evangelismos Medical Center Greece (GR) University Hospital Ghent Belgium (BE) Universitätsklinikum Münster Germany (DE) Universitätsklinikum Bonn Germany (DE) Universitätsklinikum des Saarlandes (UKS) Germany (DE) Technische Universität München (TUM) Germany (DE) Universitätsklinikum Hamburg-Eppendorf (UKE) Germany (DE) Universitätskliniken Salzburg Austria (AT) Health Sciences Research Institute of the “Germans Trias i Pujol” Foundation (IGTP) Spain (ES) Karadeniz Technical University / Karadeniz Teknik Üniversitesi (KTU) Turkey (TR) Rīga Stradiņš University Latvia (LV) Universidade Federal de Mato Grosso do Sul Brazil (BR) Public Health Institution Clinical Hospital-Shtip / Јавна Здравствена Установа Клиничка Болница - Штип Republic of North Macedonia (MK)

How to cite

APA:

Chanda, T., Haggenmueller, S., Bucher, T.C., Holland-Letz, T., Kittler, H., Tschandl, P.,... Brinker, T.J. (2025). Dermatologist-like explainable AI enhances melanoma diagnosis accuracy: eye-tracking study. Nature Communications, 16. https://doi.org/10.1038/s41467-025-59532-5

MLA:

Chanda, Tirtha, et al. "Dermatologist-like explainable AI enhances melanoma diagnosis accuracy: eye-tracking study." Nature Communications 16 (2025).

BibTeX: Download