About

Roberto Novoa is a Clinical Associate Professor in the Department of Pathology and a Clinical Assistant Professor in Dermatology at Stanford University. He is affiliated with the Center for Artificial Intelligence in Medicine & Imaging (AIMI). His work focuses on the application of artificial intelligence in medicine and imaging, contributing to research and education in these fields. As part of his role, he is involved in advancing AI-driven healthcare solutions and fostering interdisciplinary collaboration within Stanford's medical and technological communities.

Research topics

• Pathology
• Medicine
• Dermatology
• Immunology
• Cancer research
• Biology

Selected publications

• Superficial <i>CIC::DUX4</i> Sarcoma With Focal <scp>WT1</scp> Expression: Histomorphologic Clues and Immunohistochemical Pitfalls

  Journal of Cutaneous Pathology · 2026-03-31

  articleSenior author

  CIC::DUX4 sarcoma (CDS) is a rare cutaneous and deep soft tissue tumor in the differential diagnosis of small round blue cell sarcomas. CDS, by definition, is negative for rearrangements in EWSR1, and has been reported to have strong nuclear WT1 expression and ETV4 expression in most published cases so far. We herein report a case of challenging CDS diagnosed by genomic profiling, with weak, focal cytoplasmic WT-1 positivity, positive staining for CD56, CD138 (subset), and CD4 (subset). This case highlights the challenges involved in evaluating cutaneous small round blue cell tumors based on morphology and immunohistochemistry alone, especially in the presence of an atypical immunophenotype.

  PublisherDOI

• ‘Leukaemic pernio’: perniosis-like presentations of chronic lymphocytic leukaemia

  Clinical and Experimental Dermatology · 2026-01-22

  article

  Chronic lymphocytic leukemia (CLL) is the most common type of leukemia in adults. Skin involvement by malignant CLL cells, or leukemia cutis (LC), typically presents as localized or generalized papulonodular lesions. Here we report a series of five patients with CLL who presented with perniosis-like skin findings on the ears, nose, and/or digits. Histopathologic analysis of seven skin specimens from these five patients showed a relatively dense, nodular to diffuse, pan-dermal infiltrate of monomorphic lymphocytes, compatible with CLL. The typical histopathologic findings of perniosis, such as a robust perivascular and peri-eccrine infiltrate, dermal edema, and interface change, were absent. In one patient, the skin findings heralded the diagnosis of CLL, and in most patients the skin lesions tended to improve with systemic treatment of the patient's underlying CLL. We suggest the term 'leukemic pernio' for cases of leukemia cutis that clinically mimic perniosis.

  PublisherDOI

• Authors’ Reply: Enhancing AI-Driven Medical Translations: Considerations for Language Concordance

  JMIR Medical Education · 2025-04-11

  articleOpen access
  PublisherDOI

• New Diagnostics and Management for Trichodysplasia Spinulosa

  JAMA Dermatology · 2025-11-19

  article

  This case report describes a patient with a history of refractory acute myeloid leukemia who presented with white, spiny hyperkeratotic papules on his face and neck, and was subsequently diagnosed with trichodysplasia spinulosa.

  PublisherDOI

• A new prognostic index (CLIPI) for advanced cutaneous lymphoma enables precise patient risk stratification

  Blood · 2025-07-15 · 17 citations

  article

  ABSTRACT: Advanced mycosis fungoides (MF) and Sézary syndrome (SS) have a poor overall survival (OS) of <5 years. Studies have found the current staging (IA-IVB) is inadequate for risk stratification. The PROCLIPI (Prospective Cutaneous Lymphoma International Prognostic Index) study was launched in 2015 at 46 international expert MF/SS centers, prospectively collecting predefined data sets in patients with newly diagnosed MF/SS, to determine a cutaneous lymphoma IPI (CLIPI). Five hundred fifty-two patients with advanced stage MF/SS were recruited. The 5-year OS was 50.0% for stage IIB, 64.8% for stage IIIA, 43.9% for stage IIIB, 50.8% for stage IVA1, 25.9% for stage IVA2, and 36.9% for stage IVB. Factors at diagnosis associated with a significantly worse survival were N3 status (P < .001), age >60 years (P < .001), raised serum lactate dehydrogenase (P = .005), and large-cell transformation in skin (P = .006). Modeling these 4 independent risk factors into a CLIPI found that there was a worse OS in high- vs low-risk (P < .001), high- vs intermediate-risk (P = .002) and intermediate- vs low-risk (P = .010) groups. Five-year OS was 63.3%, 44.7%, and 18.3% in the low-, intermediate-, and high-risk groups, respectively. In this advanced stage cohort there was a low 5-year survival and increasing stage was not associated with worsening survival. The use of CLIPI to stratify patients into risk groups has the potential to improve outcomes and aid optimal treatment selection. This trial was registered at www.ClinicalTrials.gov as #NCT02848274.

  PublisherDOI

• Multimodal Image Dataset for AI-Based Skin Cancer (MIDAS) Benchmarking

  NEJM AI · 2025-05-20 · 5 citations

  articleSenior author
  PublisherDOI

• Metastatic Gastrointestinal Stromal Tumor to the Scalp: A Case Report Highlighting Diagnostic Challenges and Molecular Correlation

  Journal of Cutaneous Pathology · 2025-06-21 · 1 citations

  article

  Gastrointestinal stromal tumors (GISTs) are rare gastrointestinal mesenchymal neoplasms. While the liver and peritoneum are the most common metastatic sites, skin involvement is rare. Nevertheless, recognizing skin metastases is crucial as it guides targeted treatment and indicates the possibility of widespread disease. Skin metastases pose diagnostic challenges due to histopathologic and clinical variability, often resembling other tumors. We present an 81-year-old female with metastatic GIST involving the scalp with corroborative molecular data. This case underscores the fundamental nature of clinicopathologic correlation and the role of molecular analysis in aiding diagnosis, guiding treatment decisions, and revealing the prognostic implications of rare metastatic patterns of GISTs.

  PublisherDOI

• Augmented Intelligence and Dermatology – Part I: Core Concepts and Applications

  Journal of the American Academy of Dermatology · 2025-03-01 · 1 citations

  reviewSenior author
  PublisherDOI

• ImageCLEF 2025: Multimedia Retrieval in Medical, Social Media and Content Recommendation Applications

  Lecture notes in computer science · 2025-01-01 · 1 citations

  book-chapter
  PublisherDOI

• DermaVQA-DAS: Dermatology Assessment Schema (DAS) & Datasets for Closed-Ended Question Answering & Segmentation in Patient-Generated Dermatology Images

  ArXiv.org · 2025-12-30

  articleOpen access

  Recent advances in dermatological image analysis have been driven by large-scale annotated datasets; however, most existing benchmarks focus on dermatoscopic images and lack patient-authored queries and clinical context, limiting their applicability to patient-centered care. To address this gap, we introduce DermaVQA-DAS, an extension of the DermaVQA dataset that supports two complementary tasks: closed-ended question answering (QA) and dermatological lesion segmentation. Central to this work is the Dermatology Assessment Schema (DAS), a novel expert-developed framework that systematically captures clinically meaningful dermatological features in a structured and standardized form. DAS comprises 36 high-level and 27 fine-grained assessment questions, with multiple-choice options in English and Chinese. Leveraging DAS, we provide expert-annotated datasets for both closed QA and segmentation and benchmark state-of-the-art multimodal models. For segmentation, we evaluate multiple prompting strategies and show that prompt design impacts performance: the default prompt achieves the best results under Mean-of-Max and Mean-of-Mean evaluation aggregation schemes, while an augmented prompt incorporating both patient query title and content yields the highest performance under majority-vote-based microscore evaluation, achieving a Jaccard index of 0.395 and a Dice score of 0.566 with BiomedParse. For closed-ended QA, overall performance is strong across models, with average accuracies ranging from 0.729 to 0.798; o3 achieves the best overall accuracy (0.798), closely followed by GPT-4.1 (0.796), while Gemini-1.5-Pro shows competitive performance within the Gemini family (0.783). We publicly release DermaVQA-DAS, the DAS schema, and evaluation protocols to support and accelerate future research in patient-centered dermatological vision-language modeling (https://osf.io/72rp3).

  PublisherOA PDF

Frequent coauthors

• Ryanne A. Brown

  Stanford University

  31 shared

• Kerri E. Rieger

  29 shared

• Bernice Y. Kwong

  Stanford University

  19 shared

• Emily Y. Chu

  University of Pennsylvania

  18 shared

• Jennifer Y. Wang

  Stanford Medicine

  18 shared

• Justin Ko

  17 shared

• Kavita Y. Sarin

  Stanford University

  15 shared

• Susan M. Swetter

  15 shared

Education

• MD, Pathology, Dermatology

  Stanford University

• MD

  Harvard Medical School

• AB, Government

  Harvard College

  2003