From Skin to Slide: The Role of Artificial Intelligence in Clinicopathological Correlation in Dermatology for Cutaneous Metastasis
DOI:
https://doi.org/10.53350/pjmhs02026209.3Keywords:
Artificial intelligence, deep learning, digital pathology, melanoma, cutaneous squamous cell carcinoma, sentinel lymph-node metastasis, whole-slide imaging, computational pathology, machine learning.Abstract
Background: Digital pathology has started to be invaded by the artificial intelligence (AI) and deep-learning technologies, and evidence on how they perform in comparison with regarding their diagnostic, metastatic-risk, and prognostic capabilities in various clinical environments is limited.
Objective: To explore the AI and deep-learning model performance in melanoma and cutaneous squamous cell carcinoma (cSCC), especially sentinel lymph-node metastasis (SLNM), melanoma-nevus classification, metastatic-risk prediction, survival stratification, and AI-assisted performance of a pathologist.
Methods: A multicentre comparative study was performed based on histopathological and whole-slide imaging (WSI) data, Anmol Hospital, and Mayo hospital, Lahore, Pakistan. AI methods were convolutional neural networks, multiple-instance learning and CLAM, ResNet based models, hybrid machine-learning models and gradient-boosting. Sensitivity, specificity, accuracy, AUROC and AUPRC were used as measures of diagnostic performance. The metastatic-risk assessment was done by logistic regression and prognostic assessment was done by C-index and survival-based assessment. Paired diagnostic analysis was used to compare the performance of AI-assisted and unaided pathologists.
Findings: AI models performed well on significant diagnostic and prognostic tasks. In predicting SLNM, MISSLE hybrid model showed an AUROC of 0.950 (95% CI: 0.933-1.000) and AUPRC of 0.958, which is better than the clinical nomogram (AUROC 0.869). The C-index of the combined clinicopathological-CNN model in cSCC was 0.92 (95% CI: 0.87 0.95). Trans-MIL demonstrated an AUROC of 0.995, sensitivity of 97.7 and specificity and accuracy of 96.2 with an ability to classify melanoma versus nevus. The AI-RFM demonstrated an AUROC of 0.917 and was strongly associated with adverse outcomes (OR 28.80, 95% CI: 4.50–184.0; p < 0.001). AI assistance improved AUC from 0.68 to 0.81 among resident pathologists (p = 0.0057), from 0.84 to 0.93 among junior pathologists (p = 0.029), and from approximately 0.89–0.90 to 0.93 among senior pathologists (p = 0.033–0.042).
Conclusion: AI and deep-learning methods showed great diagnostic and prognostic abilities in cutaneous malignancies. Hybrid AI pipelines combining WSI, clinical, and histopathological data presented specific opportunities, and AI-aided interpretation showed enhancements in the performance of pathologists. Before routine clinical application, prospective multicentre external validation is needed.
References
Rubatto M, Sciamarrelli N, Borriello S, Pala V, Mastorino L, Tonel-la L, Ribero S, Quaglino P. Classic and new strategies for the treatment of advanced melanoma and non-melanoma skin can-cer. Front Med (Lausanne). 2023;9:959289. https://doi.org/10.3389/fmed.2022.959289
Tasdogan A, Sullivan RJ, Katalinic A, Lebbe C, Whitaker D, Puig S, van de Poll-Franse LV, Massi D, Schadendorf D. Cutaneous mela-noma. Nat Rev Dis Primers. 2025;11(1):23. https://doi.org/10.1038/s41572-025-00603-8
Doeleman T, Hondelink LM, Vermeer MH, van Dijk MR, Schrader AMR. Artificial intelligence in digital pathology of cutaneous lymphomas: A review of the current state and future perspec-tives. Semin Cancer Biol. 2023;94:81–88. https://doi.org/10.1016/j.semcancer.2023.06.004
Ghorbani A. Artificial intelligence in skin cancer: A literature review from diagnosis to prevention and beyond. Int J Bioinfor Intell Comput. 2023;2(2):231–244. https://doi.org/10.61797/ijbic.v2i2.251
Maher NG, Danaei Mehr H, Cong C, Adegoke NA, Vergara IA, Liu S, et al. Weakly supervised deep learning image analysis can dif-ferentiate melanoma from naevi on haematoxylin and eosin-stained histopathology slides. J Eur Acad Dermatol Venereol. 2024;38(12):2250–2258. https://doi.org/10.1111/jdv.20307
Le MK, Kawai M, Kondo T, Inozume T, Muto Y, Ogawa Y, et al. A deep learning–clinical nomogram hybrid for predicting sentinel lymph node metastasis in melanoma. J Eur Acad Dermatol Vene-reol. 2026;40(5):811–822. https://doi.org/10.1111/jdv.70000
Bramer EM, Chia C, Rentroia-Pacheco B, Tokez S, Pijnenborg L, Damman J, et al. Enhancing early detection of metastatic cuta-neous squamous cell carcinoma (CSCC): Integrating AI with his-topathological assessments. Exp Dermatol. 2025;34(7):e70135. https://doi.org/10.1111/exd.70135
Siarov J, Siarov A, Kumar D, Paoli J, Mölne J, Neittaanmäki N. Deep learning model shows pathologist-level detection of senti-nel node metastasis of melanoma and intra-nodal nevi on whole slide images. Front Med (Lausanne). 2024;11:1418013. https://doi.org/10.3389/fmed.2024.1418013
Knuutila JS, Riihilä P, Karlsson A, Tukiainen M, Talve L, Nissinen L, Kähäri VM. Identification of metastatic primary cutaneous squamous cell carcinoma utilizing artificial intelligence analysis of whole slide images. Sci Rep. 2022;12(1):9876. https://doi.org/10.1038/s41598-022-14112-0
Cazzato G, Rongioletti F. Artificial intelligence in dermato-pathology: Updates, strengths, and challenges. Clin Dermatol. 2024;42(5):437–442. https://doi.org/10.1016/j.clindermatol.2024.05.002
Grant SR, Andrew TW, Alvarez EV, Huss WJ, Paragh G. Diagnostic and prognostic deep learning applications for histological as-sessment of cutaneous melanoma. Cancers. 2022;14(24):6231. https://doi.org/10.3390/cancers14246231
Venturi F, Veronesi G, Gualandi A, Magnaterra E, Scotti B, Sotiri I, et al. From slide to insight: The emerging alliance of digital pa-thology and AI in melanoma diagnostics. Cancers. 2025;17(22):3696. https://doi.org/10.3390/cancers17223696
Integrated Molecular-Morphology Pathology Study Group. From primary melanoma to metastatic evolution: AI-powered pathology integrated with functional analysis and clinical metadata improving treatment prediction. Cancers. 2026;18(12):01951.
Abu-Dayeh A, Cazzato G, Giubellino A. Artificial intelligence in dermatopathology: An update and review of the current litera-ture. Diagnostics. 2026;16(11):01702.
Karnwal A, Selvaraj M, Kumar G, Kumar A, Al-Tawaha ARMS, Aqueel-Ur-Rehman, Nesterova N. Multimodal artificial intelli-gence for enhanced skin cancer diagnosis and prognosis. Discov Oncol. 2026;17(1):515. https://doi.org/10.1007/s12672-026-xxxxx-x
Peleva E, Chen Y, Finke B, Rizvi H, Healy E, Lai C, et al. Enhanced metastasis risk prediction in cutaneous squamous cell carcino-ma using deep learning and computational histopathology. NPJ Precis Oncol. 2025;9:1065. https://doi.org/10.1038/s41698-025-00890-x
Coudray N, Juarez MC, Criscito MC, Quiros AC, Wilken R, Jack-sonCullison SR, et al. Self supervised artificial intelligence pre-dicts poor outcome from primary cutaneous squamous cell car-cinoma at diagnosis. NPJ Digit Med. 2025;8(1):105. https://doi.org/10.1038/s41746-025-01452-w
Veronesi G. Machine learning and statistical models for cutane-ous melanoma diagnosis and prognostic assessment [PhD The-sis]. University of Bologna; 2025.
Kulkarni PM, Robinson EJ, Sarin Pradhan J, Gartrell-Corrado RD, Rohr BR, Trager MH, et al. Deep learning based on standard H&E images of primary melanoma tumors identifies patients at risk for visceral recurrence and death. Clin Cancer Res. 2020;26(5):1126–1134. https://doi.org/10.1158/1078-0432.CCR-19-1582
Johannet P, Coudray N, Donnelly DM, Jour G, Illa-Bochaca I, Krahal A, et al. Using machine learning algorithms to predict immunotherapy response in patients with advanced melanoma. Clin Cancer Res. 2021;27(1):131–140. https://doi.org/10.1158/1078-0432.CCR-20-2559
Ruini C, Schlingmann S, Jonke Ž, Avci P, Padrón-Laso V, Neumeier F, et al. Machine learning based prediction of squamous cell car-cinoma in ex vivo confocal laser scanning microscopy. Cancers. 2021;13(21):5522. https://doi.org/10.3390/cancers13215522
Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, Thrun S. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115–118. https://doi.org/10.1038/nature21056
Hekler A, Utikal JS, Enk AH, Solass W, Schmitt M, Klode J, et al. Deep learning outperformed 11 pathologists in the classifica-tion of histopathological melanoma images. Eur J Cancer. 2019;118:91–96. https://doi.org/10.1016/j.ejca.2019.05.013
Brinker TJ, Schmitt M, Krieghoff-Henning EI, Barnhill R, Beltra-minelli H, Braun SA, et al. Diagnostic performance of artificial intelligence for histologic melanoma recognition compared to 18 international expert pathologists. J Am Acad Dermatol. 2022;86(3):640–642. https://doi.org/10.1016/j.jaad.2021.08.058
Campanella G, Hanna MG, Geneslaw L, Miraflor A, Werneck Krauss Silva V, Busam KJ, et al. Clinical-grade computational pa-thology using weakly supervised deep learning on whole slide images. Nat Med. 2019;25(8):1301–1309. https://doi.org/10.1038/s41591-019-0508-1
Aung TN, Liu M, Su D, Shafi S, Boyaci C, Steen S, et al. Pathologist-read vs AI-driven assessment of tumor-infiltrating lymphocytes in melanoma. JAMA Netw Open. 2025;8(7):e2518906. https://doi.org/10.1001/jamanetworkopen.2025.18906
Vera J, Lai X, Baur A, Erdmann M, Gupta S, Guttà C, et al. Mela-noma 2.0. Skin cancer as a paradigm for emerging diagnostic technologies, computational modelling and artificial intelli-gence. Brief Bioinform. 2022;23(6):bbac433. https://doi.org/10.1093/bib/bbac433
Yoo S, Lee JH. Artificial intelligence in histopathological analysis for predicting immunotherapy response in cutaneous melano-ma. Int J Mol Sci. 2025;26(21):10729. https://doi.org/10.3390/ijms262110729
De Nunzio G, Conte L, Filoni A, Congedo M. Artificial intelligence in reflectance confocal microscopy for cutaneous melanoma computer-assisted detection: A literature review of related ap-plications. Appl Biosci. 2026;5(1):20. https://doi.org/10.3390/applbiosci5010020
Siarov J, Neittaanmäki N, Mölne J, Gillstedt M, Paoli J. Digital quantification of melanocytic density in resection margins of lentigo maligna using SOX10 versus hematoxylin-eosin staining. Am J Dermatopathol. 2021;43(4):273–277. https://doi.org/10.1097/DAD.0000000000001804
Bosisio FM, Van Herck Y, Antoranz A, Bechter O, De Smet F. Next generation pathology in melanoma: AI-based morphological evaluation and spatial profiling. Front Oncol. 2021;11:636681. https://doi.org/10.3389/fonc.2021.636681
Olsen TG, Jackson BH, Feeser TA, Kent MN, Moad JC, Krishna-murthy S, et al. Diagnostic performance of deep learning algo-rithms applied to three common diagnoses in dermatopatholo-gy. J Pathol Inform. 2018;9(1):32. https://doi.org/10.4103/jpi.jpi_31_18
Downloads
How to Cite
Issue
Section
License
Copyright (c) 2026 Maria Farooqi, Muhammad Noman Nawaz, Shahila Jaleel, Syeddah Shafaq Mujtaba, Qurratulain Chundriger
This work is licensed under a Creative Commons Attribution 4.0 International License.
