{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"Research Square"}],"indexed":{"date-parts":[[2026,8,17]],"date-time":"2026-08-17T21:29:18Z","timestamp":1787002158761,"version":"build-2736575974"},"posted":{"date-parts":[[2026,8,17]]},"group-title":"In Review","reference-count":26,"publisher":"Springer Science and Business Media LLC","license":[{"start":{"date-parts":[[2026,8,17]],"date-time":"2026-08-17T00:00:00Z","timestamp":1786924800000},"content-version":"unspecified","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100000048","name":"American Cancer Society","doi-asserted-by":"publisher","award":["DICRIDG-21-071-01-DICRIDG"],"award-info":[{"award-number":["DICRIDG-21-071-01-DICRIDG"]}],"id":[{"id":"10.13039\/100000048","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100006545","name":"National Institute on Minority Health and Health Disparities","doi-asserted-by":"publisher","award":["U54MD007586"],"award-info":[{"award-number":["U54MD007586"]}],"id":[{"id":"10.13039\/100006545","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,7,27]]},"abstract":"<title>Abstract<\/title>\n                <p>\n                  <bold>Background:<\/bold>\n                  Malignant melanoma (MM) is the most aggressive form of metastatic skin cancer, for which early detection is critical to reducing morbidity and mortality while improving patient outcomes. Recent advances in multimodal large language models (LLMs) have demonstrated considerable potential for melanoma detection and clinical decision support. However, their diagnostic accuracy remains inferior to that of specialized convolutional neural networks (CNNs). Whether LLMs can effectively integrate predictions from specialized CNN models to improve melanoma diagnosis has not been systematically investigated.\n                  <bold>Methods:<\/bold>\n                  We developed a hybrid diagnostic approach, termed GPT Fusion, in which GPT-5.5 serves as a reasoning-based fusion framework rather than an image classifier. Instead of directly analyzing lesion images, GPT-5.5 received structured outputs from two independently developed CNN models: a multimodal ResNet-50 model trained on the MILK10K dataset for multiclass skin lesion classificationand the first-place 90-model SIIM-ISIC ensemble optimized for melanoma detection. GPT-5.5 integrated these complementary predictions to generate the final diagnosis. The framework was evaluated using the publicly available Derm7pt dataset. Performance was assessed for melanoma prediction (melanoma vs. non-melanoma), malignancy prediction (malignant vs. benign lesions), primary diagnosis, and top-3 differential diagnosis.\n                  <bold>Results<\/bold>\n                  : GPT Fusion substantially outperformed image-based GPT-5.5 across all evaluation tasks. For melanoma prediction, accuracy improved from 63.4% to 86.9%, while ROC AUC increased from 0.750 to 0.923 and PR AUC from 0.547 to 0.853. For malignancy prediction, GPT Fusion not only exceeded the performance of image-based GPT-5.5 but also surpassed the specialized SIIM-ISIC ensemble in sensitivity (83.9%), F1 score (73.6%), ROC AUC (0.894), and PR AUC (0.823), while maintaining comparable overall accuracy. GPT Fusion further improved primary diagnosis accuracy and achieved the highest top-3 differential diagnosis accuracy (89.0%).\n                  <bold>Conclusions<\/bold>\n                  : These findings demonstrate that LLMs can provide greater clinical value as reasoning-based integrators of specialized AI systems than as standalone image classifiers. By combining the complementary strengths of specialized CNN models, GPT-5.5 substantially improved diagnostic performance while providing interpretable reasoning. This hybrid strategy offers a promising direction for developing more accurate, transparent, and clinically useful AI systems for melanoma diagnosis.\n                <\/p>","DOI":"10.21203\/rs.3.rs-10500601\/v2","type":"posted-content","created":{"date-parts":[[2026,8,17]],"date-time":"2026-08-17T20:36:27Z","timestamp":1786998987000},"source":"Crossref","is-referenced-by-count":0,"title":["GPT Fusion: Reasoning-Based Integration of Specialized Convolutional Neural Networks for Melanoma Diagnosis"],"prefix":"10.21203","author":[{"ORCID":"https:\/\/orcid.org\/0009-0007-9318-6535","authenticated-orcid":false,"given":"Katie L.","family":"Frederickson","sequence":"first","affiliation":[{"name":"Meharry Medical College"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Dong","family":"Li","sequence":"additional","affiliation":[{"name":"University of Chicago"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Oren D.","family":"Edrich","sequence":"additional","affiliation":[{"name":"Meharry Medical College"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ashish C.","family":"Bhatia","sequence":"additional","affiliation":[{"name":"Northwestern University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Samuel E.","family":"Adunyah","sequence":"additional","affiliation":[{"name":"Meharry Medical College"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5125-3724","authenticated-orcid":false,"given":"Qingguo","family":"Wang","sequence":"additional","affiliation":[{"name":"Meharry Medical College"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","reference":[{"key":"ref1","unstructured":"Key Statistics for Melanoma Skin Cancer Cancer.org: American Cancer Society (2026) [Available from: https:\/\/www.cancer.org\/cancer\/types\/melanoma-skin-cancer\/about\/key-statistics.html"},{"issue":"7639","key":"ref2","doi-asserted-by":"crossref","first-page":"115","DOI":"10.1038\/nature21056","article-title":"Dermatologist-level classification of skin cancer with deep neural networks","volume":"542","author":"Esteva A","year":"2017","unstructured":"Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM et al (2017) Dermatologist-level classification of skin cancer with deep neural networks. Nature 542(7639):115\u2013118","journal-title":"Nature"},{"key":"ref3","doi-asserted-by":"crossref","first-page":"91","DOI":"10.1016\/j.ejca.2019.06.012","article-title":"Deep learning outperformed 11 pathologists in the classification of histopathological melanoma images","volume":"118","author":"Hekler A","year":"2019","unstructured":"Hekler A, Utikal JS, Enk AH, Solass W, Schmitt M, Klode J et al (2019) Deep learning outperformed 11 pathologists in the classification of histopathological melanoma images. Eur J Cancer 118:91\u201396","journal-title":"Eur J Cancer"},{"issue":"6","key":"ref4","doi-asserted-by":"crossref","first-page":"900","DOI":"10.1038\/s41591-020-0842-3","article-title":"A deep learning system for differential diagnosis of skin diseases","volume":"26","author":"Liu Y","year":"2020","unstructured":"Liu Y, Jain A, Eng C, Way DH, Lee K, Bui P et al (2020) A deep learning system for differential diagnosis of skin diseases. Nat Med 26(6):900\u2013908","journal-title":"Nat Med"},{"issue":"1","key":"ref5","doi-asserted-by":"crossref","DOI":"10.1186\/s12880-024-01356-8","article-title":"Enhanced skin cancer diagnosis using optimized CNN architecture and checkpoints for automated dermatological lesion classification","volume":"24","author":"Musthafa MM","year":"2024","unstructured":"Musthafa MM, T R M VVK, Guluwadi S (2024) Enhanced skin cancer diagnosis using optimized CNN architecture and checkpoints for automated dermatological lesion classification. BMC Med Imaging. ;24(1)","journal-title":"BMC Med Imaging"},{"key":"ref6","doi-asserted-by":"crossref","first-page":"52","DOI":"10.1016\/j.jdin.2023.08.019","article-title":"Use of an elastic-scattering spectroscopy and artificial intelligence device in the assessment of lesions suggestive of skin cancer: A comparative effectiveness study","volume":"14","author":"Manolakos D","year":"2024","unstructured":"Manolakos D, Patrick G, Geisse JK, Rabinovitz H, Buchanan K, Hoang P et al (2024) Use of an elastic-scattering spectroscopy and artificial intelligence device in the assessment of lesions suggestive of skin cancer: A comparative effectiveness study. JAAD Int 14:52\u201358","journal-title":"JAAD Int"},{"issue":"5","key":"ref7","doi-asserted-by":"crossref","first-page":"297","DOI":"10.3390\/bios15050297","article-title":"Review of Non-Invasive Imaging Technologies for Cutaneous Melanoma","volume":"15","author":"Horton L","year":"2025","unstructured":"Horton L, Fakhoury JW, Manwar R, Rajabi-Estarabadi A, Turk D, O\u2019Leary S et al (2025) Review of Non-Invasive Imaging Technologies for Cutaneous Melanoma. Biosensors 15(5):297","journal-title":"Biosensors"},{"issue":"1","key":"ref8","doi-asserted-by":"crossref","DOI":"10.1038\/s44401-025-00016-5","article-title":"Trust in AI-assisted health systems and AI\u2019s trust in humans","volume":"2","author":"Sagona M","year":"2025","unstructured":"Sagona M, Dai T, Macis M, Darden M (2025) Trust in AI-assisted health systems and AI\u2019s trust in humans. npj Health Syst. ;2(1)","journal-title":"npj Health Syst"},{"issue":"8","key":"ref9","doi-asserted-by":"crossref","DOI":"10.1007\/s00403-024-03249-w","article-title":"Wait times for scheduling appointments with hospital affiliated dermatologists in New York City","volume":"316","author":"Basch CH","year":"2024","unstructured":"Basch CH, Hillyer GC, Gold B, Basch CE (2024) Wait times for scheduling appointments with hospital affiliated dermatologists in New York City. Arch Dermatol Res. ;316(8)","journal-title":"Arch Dermatol Res"},{"key":"ref10","unstructured":"OpenAI GPT-5 System Card2025. Available from: https:\/\/cdn.openai.com\/gpt-5-system-card.pdf"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.48550\/arxiv.2405.09818","author":"Team C","year":"2024","unstructured":"Team C, Chameleon (2024) 5\/16\/2024 Mixed-Modal Early-Fusion Foundation Models. arXiv [Internet]. Available from: https:\/\/dx.doi.org\/10.48550\/arxiv.2405.09818"},{"key":"ref12","unstructured":"Gemini https:\/\/gemini.google.com\/app: Google; 2026 [Available from: https:\/\/gemini.google.com\/app"},{"key":"ref13","doi-asserted-by":"crossref","first-page":"e88407","DOI":"10.2196\/88407","article-title":"Evaluation of GPT-5 in Periodontitis Staging and Grading: Retrospective Observational Study","volume":"10","author":"Amugo I","year":"2026","unstructured":"Amugo I, Frederickson KL, Rajakaruna H, Xie H, Gangula P, Shanker A et al (2026) Evaluation of GPT-5 in Periodontitis Staging and Grading: Retrospective Observational Study. JMIR Formative Res 10:e88407\u2013e","journal-title":"JMIR Formative Res"},{"issue":"23","key":"ref14","doi-asserted-by":"crossref","first-page":"3052","DOI":"10.3390\/diagnostics15233052","article-title":"Evaluating GPT-5 for Melanoma Detection Using Dermoscopic Images","volume":"15","author":"Wang Q","year":"2025","unstructured":"Wang Q, Amugo I, Rajakaruna H, Irudayam MJ, Xie H, Shanker A et al (2025) Evaluating GPT-5 for Melanoma Detection Using Dermoscopic Images. Diagnostics 15(23):3052","journal-title":"Diagnostics"},{"key":"ref15","article-title":"Artificial Intelligence Use in Acne Diagnosis and Management\u2014A Scoping Review","author":"Frederickson KL","year":"2025","unstructured":"Frederickson KL, Gui H, Barbieri JS, Daneshjou R (2025) Artificial Intelligence Use in Acne Diagnosis and Management\u2014A Scoping Review. Int J Dermatol","journal-title":"Int J Dermatol"},{"issue":"5","key":"ref16","doi-asserted-by":"crossref","first-page":"1057","DOI":"10.1016\/j.jaad.2023.12.062","article-title":"Can ChatGPT vision diagnose melanoma? An exploratory diagnostic accuracy study","volume":"90","author":"Shifai N","year":"2024","unstructured":"Shifai N, Van Doorn R, Malvehy J, Sangers TE (2024) Can ChatGPT vision diagnose melanoma? An exploratory diagnostic accuracy study. J Am Acad Dermatol 90(5):1057\u20131059","journal-title":"J Am Acad Dermatol"},{"key":"ref17","doi-asserted-by":"crossref","first-page":"e59273","DOI":"10.2196\/59273","article-title":"Claude 3 Opus and ChatGPT With GPT-4 in Dermoscopic Image Analysis for Melanoma Diagnosis: Comparative Performance Analysis","volume":"12","author":"Liu X","year":"2024","unstructured":"Liu X, Duan C, Kim M-K, Zhang L, Jee E, Maharjan B et al (2024) Claude 3 Opus and ChatGPT With GPT-4 in Dermoscopic Image Analysis for Melanoma Diagnosis: Comparative Performance Analysis. JMIR Med Inf 12:e59273","journal-title":"JMIR Med Inf"},{"key":"ref18","doi-asserted-by":"crossref","first-page":"e67551","DOI":"10.2196\/67551","article-title":"Evaluating the Diagnostic Accuracy of ChatGPT-4 Omni and ChatGPT-4 Turbo in Identifying Melanoma: Comparative Study","volume":"8","author":"Sattler SS","year":"2025","unstructured":"Sattler SS, Chetla N, Chen M, Hage TR, Chang J, Guo WY et al (2025) Evaluating the Diagnostic Accuracy of ChatGPT-4 Omni and ChatGPT-4 Turbo in Identifying Melanoma: Comparative Study. JMIR Dermatology 8:e67551\u2013e","journal-title":"JMIR Dermatology"},{"key":"ref19","doi-asserted-by":"crossref","first-page":"58","DOI":"10.1016\/j.jdin.2025.08.008","article-title":"Diagnostic performance of multimodal large language models in distinguishing melanoma from nevi in clinical and dermoscopic images","volume":"23","author":"Boostani M","year":"2025","unstructured":"Boostani M, Lallas A, Goldust M, N\u00e1dudvari N, L\u0151rincz K, B\u00e1nv\u00f6lgyi A et al (2025) Diagnostic performance of multimodal large language models in distinguishing melanoma from nevi in clinical and dermoscopic images. JAAD Int 23:58\u201360","journal-title":"JAAD Int"},{"key":"ref20","doi-asserted-by":"crossref","DOI":"10.3389\/fmed.2026.1816102","article-title":"Evaluation of GPT-5.2 for melanoma detection across skin tones","volume":"13","author":"Frederickson KL","year":"2026","unstructured":"Frederickson KL, Adunyah SE, Wang Q (2026) Evaluation of GPT-5.2 for melanoma detection across skin tones. Front Med. ;13","journal-title":"Front Med"},{"key":"ref21","article-title":"MILK10k: A Hierarchical Multimodal Imaging-Learning Toolkit for Diagnosing Pigmented and Nonpigmented Skin Cancer and its Simulators","author":"Tschandl P","year":"2025","unstructured":"Tschandl P, Akay BN, Rosendahl C, Rotemberg V, Todorovska V, Weber J et al (2025) MILK10k: A Hierarchical Multimodal Imaging-Learning Toolkit for Diagnosing Pigmented and Nonpigmented Skin Cancer and its Simulators. J Invest Dermatology","journal-title":"J Invest Dermatology"},{"key":"ref22","author":"Qishen Ha BL","year":"2020","unstructured":"Qishen Ha BL, Fuxu Liu (2020). Identifying Melanoma Images using EfficientNet Ensemble: Winning Solution to the SIIM-ISIC Melanoma Classification Challenge. arXiv"},{"issue":"2","key":"ref23","doi-asserted-by":"crossref","first-page":"538","DOI":"10.1109\/JBHI.2018.2824327","article-title":"Seven-Point Checklist and Skin Lesion Classification Using Multitask Multimodal Neural Nets","volume":"23","author":"Kawahara J","year":"2019","unstructured":"Kawahara J, Daneshvar S, Argenziano G, Hamarneh G (2019) Seven-Point Checklist and Skin Lesion Classification Using Multitask Multimodal Neural Nets. IEEE J Biomedical Health Inf 23(2):538\u2013546","journal-title":"IEEE J Biomedical Health Inf"},{"issue":"9","key":"ref24","doi-asserted-by":"crossref","first-page":"1718","DOI":"10.1111\/jdv.19221","article-title":"Dermoscopy of melanoma according to different body sites: Head and neck, trunk, limbs, nail, mucosal and acral","volume":"37","author":"Longo C","year":"2023","unstructured":"Longo C, Pampena R, Moscarella E, Chester J, Starace M, Cinotti E et al (2023) Dermoscopy of melanoma according to different body sites: Head and neck, trunk, limbs, nail, mucosal and acral. J Eur Acad Dermatol Venereol 37(9):1718\u20131730","journal-title":"J Eur Acad Dermatol Venereol"},{"issue":"6552","key":"ref25","doi-asserted-by":"crossref","first-page":"284","DOI":"10.1126\/science.abg1834","article-title":"Beware explanations from AI in health care","volume":"373","author":"Babic B","year":"2021","unstructured":"Babic B, Gerke S, Evgeniou T, Cohen IG (2021) Beware explanations from AI in health care. Science 373(6552):284\u2013286","journal-title":"Science"},{"issue":"11","key":"ref26","doi-asserted-by":"crossref","first-page":"e745","DOI":"10.1016\/S2589-7500(21)00208-9","article-title":"The false hope of current approaches to explainable artificial intelligence in health care","volume":"3","author":"Ghassemi M","year":"2021","unstructured":"Ghassemi M, Oakden-Rayner L, Beam AL (2021) The false hope of current approaches to explainable artificial intelligence in health care. Lancet Digit Health 3(11):e745\u2013e50","journal-title":"Lancet Digit Health"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/www.researchsquare.com\/article\/rs-10500601\/v2","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.researchsquare.com\/article\/rs-10500601\/v2.html","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,17]],"date-time":"2026-08-17T20:36:45Z","timestamp":1786999005000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.researchsquare.com\/article\/rs-10500601\/v2"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,17]]},"references-count":26,"URL":"https:\/\/doi.org\/10.21203\/rs.3.rs-10500601\/v2","relation":{"is-version-of":[{"id-type":"doi","id":"10.21203\/rs.3.rs-10500601\/v1","asserted-by":"subject"}],"has-version":[{"id-type":"doi","id":"10.21203\/rs.3.rs-10500601\/v1","asserted-by":"object"}]},"subject":[],"published":{"date-parts":[[2026,8,17]]},"subtype":"preprint"}}