{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T01:51:33Z","timestamp":1780451493191,"version":"3.54.1"},"posted":{"date-parts":[[2025,6,9]]},"group-title":"Genetic and Genomic Medicine","reference-count":24,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2025,6,9]],"date-time":"2025-06-09T00:00:00Z","timestamp":1749427200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/publicdomain\/zero\/1.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,3,12]]},"abstract":"<jats:title>ABSTRACT<\/jats:title>\n                <jats:sec>\n                  <jats:title>Objective<\/jats:title>\n                  <jats:p>To assess whether saliency-based explainable artificial intelligence (XAI) can improve diagnostic accuracy, confidence, and perceived usefulness for recognition of genetic conditions based on facial images.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Materials and methods<\/jats:title>\n                  <jats:p>Forty-four medical geneticists, divided into AI-only (n=23) and XAI-supported (n=21) groups, assessed eighteen images of individuals with and without genetic conditions. Geneticists\u2019 diagnostic accuracy and confidence were recorded before and after viewing an AI classifier\u2019s prediction probability, with or without XAI explanations. Mediation analyses were conducted to interpret geneticist behavior.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>AI-only and XAI-support improved geneticist accuracy for correct AI classifications, with mean improvements of 0.20\u00b10.13 (AI-only) and 0.19\u00b10.13 (XAI). Incorrect AI classifications decreased accuracy in both groups: \u22120.20\u00b10.22 (AI-only) and \u22120.21\u00b10.23 (XAI). Average confidence increased with correct classification: 0.30\u00b10.26 (AI-only) and 0.38\u00b10.71 (XAI) and decreased with incorrect classification: \u22120.14\u00b10.64 (AI-only) and - 0.14\u00b10.76 (XAI). AI prediction probability were generally considered useful (0.58\u00b11.1 (AI-only) and 0.72\u00b11.3 (XAI)); XAI explanations were viewed less favorably: \u22120.14\u00b11.3 (saliency maps); \u22120.19\u00b11.3 (region relevance scores). For incorrect AI classifications, there was a negative correlation between accuracy improvement and perceived AI usefulness (Spearman\u2019s Rho \u22120.31 (AI-only) and \u22120.40 (XAI)). Mediation analyses demonstrated that when AI is correct (without XAI), there was a significant mediated effect (0.583, 95% CI [0.044, 1.144]) by the model probability between user confidence and decision to follow AI.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Conclusions<\/jats:title>\n                  <jats:p>The lack of accuracy or confidence improvements together with additional qualitative responses indicate that participants did not integrate saliency-based XAI into their decisions. AI prediction probability had a greater impact on participants\u2019 decision making.<\/jats:p>\n                <\/jats:sec>","DOI":"10.1101\/2025.06.08.25328588","type":"posted-content","created":{"date-parts":[[2025,6,9]],"date-time":"2025-06-09T13:00:12Z","timestamp":1749474012000},"source":"Crossref","is-referenced-by-count":1,"title":["Application of deep learning and explainable AI-supported medical decision-making for facial phenotyping in genetic syndromes"],"prefix":"10.64898","author":[{"given":"\u00d6mer","family":"S\u00fcmer","sequence":"first","affiliation":[{"name":"Human-Centered Artificial Intelligence, University of Augsburg, Augsburg, Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tobias","family":"Huber","sequence":"additional","affiliation":[{"name":"Technische Hochschule Ingolstadt, Ingolstadt, Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-1057-1396","authenticated-orcid":false,"given":"Jiayi","family":"Cheng","sequence":"additional","affiliation":[{"name":"Medical Genetics Branch, National Human Genome Research Institute, Bethesda, MD, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dat","family":"Duong","sequence":"additional","affiliation":[{"name":"Medical Genetics Branch, National Human Genome Research Institute, Bethesda, MD, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Suzanna E.","family":"Ledgister Hanchard","sequence":"additional","affiliation":[{"name":"Medical Genetics Branch, National Human Genome Research Institute, Bethesda, MD, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Cristina","family":"Conati","sequence":"additional","affiliation":[{"name":"Department of Computer Science, University of British Columbia, BC, Canada"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Elisabeth","family":"Andr\u00e9","sequence":"additional","affiliation":[{"name":"Human-Centered Artificial Intelligence, University of Augsburg, Augsburg, Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8826-1023","authenticated-orcid":false,"given":"Benjamin D","family":"Solomon","sequence":"additional","affiliation":[{"name":"Medical Genetics Branch, National Human Genome Research Institute, Bethesda, MD, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8635-7542","authenticated-orcid":false,"given":"Rebekah L.","family":"Waikel","sequence":"additional","affiliation":[{"name":"Medical Genetics Branch, National Human Genome Research Institute, Bethesda, MD, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026031600450610000_2025.06.08.25328588v2.1","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-021-01010-x"},{"key":"2026031600450610000_2025.06.08.25328588v2.2","doi-asserted-by":"publisher","DOI":"10.1111\/cge.12716"},{"key":"2026031600450610000_2025.06.08.25328588v2.3","doi-asserted-by":"publisher","DOI":"10.1001\/jamanetworkopen.2024.40969"},{"key":"2026031600450610000_2025.06.08.25328588v2.4","doi-asserted-by":"publisher","DOI":"10.1001\/jamanetworkopen.2021.1276"},{"issue":"3","key":"2026031600450610000_2025.06.08.25328588v2.5","doi-asserted-by":"crossref","first-page":"e211276","DOI":"10.1001\/jamanetworkopen.2021.1276","article-title":"Association of clinician diagnostic performance with machine learning\u2013based decision support systems: a systematic review","volume":"4","year":"2021","journal-title":"JAMA network open"},{"key":"2026031600450610000_2025.06.08.25328588v2.6","doi-asserted-by":"publisher","DOI":"10.1093\/jamia\/ocz229"},{"key":"2026031600450610000_2025.06.08.25328588v2.7","doi-asserted-by":"crossref","first-page":"825565","DOI":"10.3389\/frai.2022.825565","article-title":"Ganterfactual\u2014counterfactual explanations for medical non-experts using generative adversarial learning","volume":"5","year":"2022","journal-title":"Frontiers in artificial intelligence"},{"key":"2026031600450610000_2025.06.08.25328588v2.8","unstructured":"Supporting self-reflection at scale with large language models: Insights from randomized field experiments in classrooms. Proceedings of the eleventh ACM conference on learning@ scale; 2024."},{"key":"2026031600450610000_2025.06.08.25328588v2.9","unstructured":"Yang J. Fast treeshap: Accelerating shap value computation for trees. arXiv preprint arXiv:2109.09847 2021."},{"issue":"10","key":"2026031600450610000_2025.06.08.25328588v2.10","doi-asserted-by":"crossref","first-page":"867","DOI":"10.1038\/s42256-022-00536-x","article-title":"Benchmarking saliency methods for chest X-ray interpretation","volume":"4","year":"2022","journal-title":"Nature Machine Intelligence"},{"key":"2026031600450610000_2025.06.08.25328588v2.11","unstructured":"Region-based saliency explanations on the recognition of facial genetic syndromes. Machine Learning for Healthcare Conference; 2023. PMLR."},{"key":"2026031600450610000_2025.06.08.25328588v2.12","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pgen.1011168"},{"key":"2026031600450610000_2025.06.08.25328588v2.13","unstructured":"Ancona M , Ceolini E , \u00d6ztireli C , Gross M. Towards better understanding of gradient-based attribution methods for deep neural networks. arXiv preprint arXiv:1711.06104 2017."},{"key":"2026031600450610000_2025.06.08.25328588v2.14","doi-asserted-by":"publisher","DOI":"10.3233\/SHTI230975"},{"key":"2026031600450610000_2025.06.08.25328588v2.15","doi-asserted-by":"crossref","first-page":"103571","DOI":"10.1016\/j.artint.2021.103571","article-title":"Local and global explanations of agent behavior: Integrating strategy summaries with saliency maps","volume":"301","year":"2021","journal-title":"Artificial Intelligence"},{"key":"2026031600450610000_2025.06.08.25328588v2.16","doi-asserted-by":"publisher","DOI":"10.1101\/2023.07.26.23293119"},{"key":"2026031600450610000_2025.06.08.25328588v2.17","doi-asserted-by":"publisher","DOI":"10.1001\/jamanetworkopen.2024.2609"},{"key":"2026031600450610000_2025.06.08.25328588v2.18","doi-asserted-by":"publisher","DOI":"10.1016\/j.xhgg.2021.100053"},{"key":"2026031600450610000_2025.06.08.25328588v2.19","doi-asserted-by":"publisher","DOI":"10.1007\/s00247-023-05789-1"},{"key":"2026031600450610000_2025.06.08.25328588v2.20","doi-asserted-by":"publisher","DOI":"10.3389\/fgene.2022.864092"},{"key":"2026031600450610000_2025.06.08.25328588v2.21","doi-asserted-by":"crossref","first-page":"990604","DOI":"10.3389\/fmed.2022.990604","article-title":"Acceptance of clinical artificial intelligence among physicians and medical students: a systematic review with cross-sectional survey","volume":"9","year":"2022","journal-title":"Frontiers in medicine"},{"key":"2026031600450610000_2025.06.08.25328588v2.22","unstructured":"Concept bottleneck models. International conference on machine learning; 2020. PMLR."},{"issue":"2","key":"2026031600450610000_2025.06.08.25328588v2.23","doi-asserted-by":"crossref","first-page":"e1011168","DOI":"10.1371\/journal.pgen.1011168","article-title":"Comparison of clinical geneticist and computer visual attention in assessing genetic conditions","volume":"20","year":"2024","journal-title":"PLoS Genetics"},{"key":"2026031600450610000_2025.06.08.25328588v2.24","doi-asserted-by":"crossref","unstructured":"Vaccaro M , Almaatouq A , Malone T . When combinations of humans and AI are useful: A systematic review and meta-analysis. Nature Human Behaviour 2024:1\u201311.","DOI":"10.31234\/osf.io\/mk46p"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.1101\/2025.06.08.25328588","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,16]],"date-time":"2026-03-16T07:45:16Z","timestamp":1773647116000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.1101\/2025.06.08.25328588"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,6,9]]},"references-count":24,"URL":"https:\/\/doi.org\/10.1101\/2025.06.08.25328588","relation":{},"subject":[],"published":{"date-parts":[[2025,6,9]]},"subtype":"preprint"}}