Stanford Advisors


All Publications


  • Foundation Model for Predicting Prognosis and Adjuvant Therapy Benefit From Digital Pathology in GI Cancers. Journal of clinical oncology : official journal of the American Society of Clinical Oncology Wang, X., Jiang, Y., Yang, S., Wang, F., Zhang, X., Wang, W., Chen, Y., Wu, X., Xiang, J., Li, Y., Jiang, X., Yuan, W., Zhang, J., Yu, K. H., Ward, R. L., Hawkins, N., Jonnagaddala, J., Li, G., Li, R. 2025: JCO2401501

    Abstract

    Artificial intelligence (AI) holds significant promise for improving cancer diagnosis and treatment. Here, we present a foundation AI model for prognosis prediction on the basis of standard hematoxylin and eosin-stained histopathology slides.In this multinational cohort study, we developed AI models to predict prognosis from histopathology images of patients with GI cancers. First, we trained a foundation model using over 130 million patches from 104,876 whole-slide images on the basis of self-supervised learning. Second, we fine-tuned deep learning models for predicting survival outcomes and validated them across seven cohorts, including 1,619 patients with gastric and esophageal cancers and 2,594 patients with colorectal cancer. We further assessed the model for predicting survival benefit from adjuvant chemotherapy.The AI models predicted disease-free survival and disease-specific survival with a concordance index of 0.726-0.797 for gastric cancer and 0.714-0.757 for colorectal cancer in the validation cohorts. The models stratified patients into high-risk and low-risk groups, with 5-year survival rates of 49%-52% versus 76%-92% in gastric cancer and 43%-72% versus 81%-98% in colorectal cancer. In multivariable analysis, the AI risk scores remained an independent prognostic factor after adjusting for clinicopathologic variables. Compared with stage alone, an integrated model consisting of stage and image information improved prognosis prediction across all validation cohorts. Finally, adjuvant chemotherapy was associated with improved survival in the high-risk group but not in the low-risk group (treatment-model interaction P = .01 and .006) for stage II/III gastric and colorectal cancer, respectively.The pathology foundation model can accurately predict survival outcomes and complement clinicopathologic factors in GI cancers. Pending prospective validation, it may be used to improve risk stratification and inform personalized adjuvant therapy.

    View details for DOI 10.1200/JCO-24-01501

    View details for PubMedID 40168636

  • A vision-language foundation model for precision oncology. Nature Xiang, J., Wang, X., Zhang, X., Xi, Y., Eweje, F., Chen, Y., Li, Y., Bergstrom, C., Gopaulchan, M., Kim, T., Yu, K. H., Willens, S., Olguin, F. M., Nirschl, J. J., Neal, J., Diehn, M., Yang, S., Li, R. 2025

    Abstract

    Clinical decision-making is driven by multimodal data, including clinical notes and pathological characteristics. Artificial intelligence approaches that can effectively integrate multimodal data hold significant promise in advancing clinical care1,2. However, the scarcity of well-annotated multimodal datasets in clinical settings has hindered the development of useful models. In this study, we developed the Multimodal transformer with Unified maSKed modeling (MUSK), a vision-language foundation model designed to leverage large-scale, unlabelled, unpaired image and text data. MUSK was pretrained on 50 million pathology images from 11,577 patients and one billion pathology-related text tokens using unified masked modelling. It was further pretrained on one million pathology image-text pairs to efficiently align the vision and language features. With minimal or no further training, MUSK was tested in a wide range of applications and demonstrated superior performance across 23 patch-level and slide-level benchmarks, including image-to-text and text-to-image retrieval, visual question answering, image classification and molecular biomarker prediction. Furthermore, MUSK showed strong performance in outcome prediction, including melanoma relapse prediction, pan-cancer prognosis prediction and immunotherapy response prediction in lung and gastro-oesophageal cancers. MUSK effectively combined complementary information from pathology images and clinical reports and could potentially improve diagnosis and precision in cancer therapy.

    View details for DOI 10.1038/s41586-024-08378-w

    View details for PubMedID 39779851

    View details for PubMedCentralID 9586871

  • Glomerular and Nephron Size and Kidney Disease Outcomes: A Comparison of Manual Versus Deep Learning Methods in Kidney Pathology. Kidney medicine Jagtap, J. M., Janowczyk, A. R., Chen, Y., Shaik, A. A., Mullan, A. F., Erickson, B. J., Sharma, V., Kline, T. L., Barisoni, L., Denic, A., Rule, A. D. 2025; 7 (1): 100939

    View details for DOI 10.1016/j.xkme.2024.100939

    View details for PubMedID 39807248

  • Clinical Relevance of Computational Pathology Analysis of Interplay Between Kidney Microvasculature and Interstitial Microenvironment. Clinical journal of the American Society of Nephrology : CJASN Chen, Y., Wang, B., Demeke, D., Fan, F., Berthier, C., Mariani, L., Lafata, K., Holzman, L., Hodgin, J., Janowczyk, A., Barisoni, L., Madabhushi, A. 2024

    Abstract

    Interstitial fibrosis and tubular atrophy (IFTA), and density and shape of peritubular capillaries (PTCs), are independently prognostic of disease progression. This study aimed to identify novel digital biomarkers of disease progression and assess the clinical relevance of the interplay between a variety of PTC characteristics and their microenvironment in glomerular diseases.A total of 344 NEPTUNE/CureGN participants were included: 112 minimal change disease, 134 focal segmental glomerulosclerosis, 61 membranous nephropathy, and 37 IgA nephropathy. A PAS-stained whole slide image per patient was manually segmented for cortex, pre- and mature IFTA. Interstitial fractional space (IFS) was computationally quantified. A deep-learning model was applied to segment PTCs. Spatial and shape PTC pathomic features (230) were extracted from the cortex, IFTA, and non-IFTA sub-regions. Participants were divided into training and testing datasets (1:1). Univariate models incorporating IFTA subregions, and IFS-PTC density were constructed. LASSO regression models were used to identify the top PTC features associated with disease progression stratified by IFTA and non-IFTA sub-regions. Machine learning models were built using the top PTC features in IFTA and non-IFTA sub-regions to prognosticate disease progression.PTC density in pre+mature IFTA and IFS, shape features in pre+mature IFTA, and spatial architecture features in the non-IFTA regions associated with disease progression. The machine learning generated risk scores showed a significant association with disease progression on the independent testing set.We uncovered previously underrecognized digital biomarkers of disease progression and the clinical relevance of the complex interplay between the status of the PTCs and the interstitial microenvironment.

    View details for DOI 10.2215/CJN.0000000597

    View details for PubMedID 39714939

  • Computational Characterization of Arteries/Arterioles in FSGS/Minimal Change Disease Zhou, J., Demeke, D. S., Li, X., Dinh, T. A., O'Connor, C., Liu, Q., Zee, J., Chen, Y., Janowczyk, A., Holzman, L. B., Mariani, L. H., Bitzer, M., Barisoni, L., Hodgin, J. B., Lafata, K. AMER SOC NEPHROLOGY. 2024
  • Computationally Derived Characterization of Tubular Changes in Relation to the Development of Interstitial Fibrosis Fan, F., Liu, Q., Zee, J., Ozeki, T., Demeke, D. S., Wang Bangchen, Jacobs, J., Shah, M. P., Farris, A., Mariani, L. H., Lafata, K., Chen Yijiang, Holzman, L. B., Hodgin, J. B., Madabhushi, A., Barisoni, L., Janowczyk, A. AMER SOC NEPHROLOGY. 2024
  • Mesoscopic structure graphs for interpreting uncertainty in non-linear embeddings. Computers in biology and medicine Zhao, J., Liu, X., Tang, H., Wang, X., Yang, S., Liu, D., Chen, Y., Chen, Y. V. 2024; 182: 109105

    Abstract

    Probabilistic-based non-linear dimensionality reduction (PB-NL-DR) methods, such as t-SNE and UMAP, are effective in unfolding complex high-dimensional manifolds, allowing users to explore and understand the structural patterns of data. However, due to the trade-off between global and local structure preservation and the randomness during computation, these methods may introduce false neighborhood relationships, known as distortion errors and misleading visualizations. To address this issue, we first conduct a detailed survey to illustrate the design space of prior layout enrichment visualizations for interpreting DR results, and then propose a node-link visualization technique, ManiGraph. This technique rethinks the neighborhood fidelity between the high- and low-dimensional spaces by constructing dynamic mesoscopic structure graphs and measuring region-adapted trustworthiness. ManiGraph also addresses the overplotting issue in scatterplot visualization for large-scale datasets and supports examining in unsupervised scenarios. We demonstrate the effectiveness of ManiGraph in different analytical cases, including generic machine learning using 3D toy data illustrations and fashion-MNIST, a computational biology study using a single-cell RNA sequencing dataset, and a deep learning-enabled colorectal cancer study with histopathology-MNIST.

    View details for DOI 10.1016/j.compbiomed.2024.109105

    View details for PubMedID 39265479

  • Clinical Relevance of Computationally Derived Tubular Features: Spatial Relationships and the Development of Tubulointerstitial Scarring in MCD/FSGS. medRxiv : the preprint server for health sciences Fan, F., Liu, Q., Zee, J., Ozeki, T., Demeke, D., Yang, Y., Farris, A. B., Wang, B., Shah, M., Jacobs, J., Mariani, L., Lafata, K., Rubin, J., Chen, Y., Holzman, L., Hodgin, J. B., Madabhushi, A., Barisoni, L., Janowczyk, A. 2024

    Abstract

    Background: Visual scoring of tubular damage has limitations in capturing the full spectrum of structural changes and prognostic potential. We investigate if computationally quantified tubular features can enhance prognostication and reveal spatial relationships with interstitial fibrosis.Methods: Deep-learning and image-processing-based segmentations were employed in N=254/266 PAS-WSIs from the NEPTUNE/CureGN datasets (135/153 focal segmental glomerulosclerosis and 119/113 minimal change disease) for: cortex, tubular lumen (TL), epithelium (TE), nuclei (TN), and basement membrane (TBM). N=104 pathomic features were extracted from these segmented tubular substructures and summarized at the patient level using summary statistics. The tubular features were quantified across the biopsy and in manually segmented regions of mature interstitial fibrosis and tubular atrophy (IFTA), pre-IFTA and non-IFTA in the NEPTUNE dataset. Minimum Redundancy Maximum Relevance was used in the NEPTUNE dataset to select features most associated with disease progression and proteinuria remission. Ridge-penalized Cox models evaluated their predictive discrimination compared to clinical/demographic data and visual-assessment. Models were evaluated in the CureGN dataset.Results: N=9 features were predictive of disease progression and/or proteinuria remission. Models with tubular features had high prognostic accuracy in both NEPTUNE and CureGN datasets and increased prognostic accuracy for both outcomes (5.6%-7.7% and 1.6%-4.6% increase for disease progression and proteinuria remission, respectively) compared to conventional parameters alone in the NEPTUNE dataset. TBM thickness/area and TE simplification progressively increased from non- to pre- and mature IFTA.Conclusions: Previously under-recognized, quantifiable, and clinically relevant tubular features in the kidney parenchyma can enhance understanding of mechanisms of disease progression and risk stratification.

    View details for DOI 10.1101/2024.07.19.24310619

    View details for PubMedID 39072032

  • The chromatin landscape of healthy and injured cell types in the human kidney NATURE COMMUNICATIONS Gisch, D. L., Brennan, M., Lake, B. B., Basta, J., Keller, M. S., Melo Ferreira, R., Akilesh, S., Ghag, R., Lu, C., Cheng, Y., Collins, K. S., Parikh, S. V., Rovin, B. H., Robbins, L., Stout, L., Conklin, K. Y., Diep, D., Zhang, B., Knoten, A., Barwinska, D., Asghari, M., Sabo, A. R., Ferkowicz, M. J., Sutton, T. A., Kelly, K. J., De Boer, I. H., Rosas, S. E., Kiryluk, K., Hodgin, J. B., Alakwaa, F., Winfree, S., Jefferson, N., Tuerkmen, A., Gaut, J. P., Gehlenborg, N., Phillips, C. L., El-Achkar, T. M., Dagher, P. C., Hato, T., Zhang, K., Himmelfarb, J., Kretzler, M., Mollah, S., Lake, B., Morales, A., Stillman, I., Lecker, S., Bogen, S., Verma, A., Yu, G., Schmidt, I., Henderson, J., Beck, L., Yadati, P., Waikar, S., Amodu, A. A., Maikhor, S., Ilori, T., Colona, M. R., Weins, A., Mcmahon, G., Hacohen, N., Greka, A., Marshall, J. L., Hoover, P. J., Viswanathan, V. S., Crawford, D., Aulisio, M., Bush, W., Chen, Y., Madabhushi, A., O'Malley, C., Gadegbeku, C., Sendrey, D., Poggio, E., O'Toole, J., Sedor, J., Taliercio, J., Bush, L., Herlitz, L., Palmer, E., Nguyen, J., Spates-Harden, K., Cooperman, L., Jolly, S., Vinovskis, C., Bomback, A., Barasch, J., Kiryluk, K., Appelbaum, P., D'Agati, V., Berrouet, C., Mehl, K., Sabatello, M., Shang, N., Balderes, O., Canetta, P. A., Kudose, S., de Pinho Goncalves, J., Migas, L., van de Plas, R., Lardenoije, R., Barisoni, L., Rennke, H., Verdoes, A., Sabo, A., Gisch, D., Williams, J., Kelly, K., Dunn, K., Eadon, M., Ferkowicz, M., Dagher, P., Winfree, S., Bledsoe, S., Wofford, S., (El-Achkar), T., Sutton, T., Bowen, W., Slade, A., Record, E., Cheng, Y., Jain, Y., Herr, B., Quardokus, E., Wang, A., Villalobos, C., Parikh, C., Atta, M., Menez, S., Wen, Y., Xu, A., Bernard, L., Johansen, C., Chen, S., Rosas, S., Donohoe, I., Sun, J., Knight, R., Shpigel, A., Bebiak, J., Saul, J., Ardayfio, J., Koewler, R., Pinkeney, R., Campbell, T., Azeloglu, E., Nadkarni, G., He, J., Tokita, J., Campbell, K., Patel, M., Lefferts, S., Iyengar, S., Ward, S., Coca, S., He, C., Xiong, Y., Prasad, P., Rovin, B., Shapiro, J. P., Parikh, S., Madhavan, S. M., Lukowski, J., Velickovic, D., Pasa-Tolic, L., Oliver, G., Troyanskaya, O., Sealfon, R., Mao, W., Wong, A., Pollack, A., Goltsev, Y., Ginley, B., Lutnick, B., Nolan, G., Anjani, K., Mukatash, T., Laszik, Z. G., Campos, B., Thajudeen, B., Beyda, D., Bracamonte, E., Brosius, F., Woodhead, G., Mendoza, K., Marquez, N., Scott, R., Tsosie, R., Saunders, M., Rike, A., Woodle, E., Lee, P. J., Alloway, R. R., Shi, T., Hsieh, E., Kendrick, J., Thurman, J., Wrobel, J., Pyle, L., Bjornstad, P., Lucarelli, N., Sarder, P., Renteria, A., Ricardo, A., Srivastava, A., Redmond, D., Carmona-Powell, E., Bui, J., Lash, J., Fox, M., Meza, N., Gaba, R., Setty, S., Kelly, T., Lienczewski, C., Demeke, D., Otto, E., Ascani, H., Hodgin, J., Schaub, J., Hartman, J., Mariani, L., Bitzer, M., Rose, M., Bonevich, N., Conser, N., Mccown, P., Dull, R., Menon, R., Reamy, R., Eddy, S., Balis, U., Blanc, V., Nair, V., He, Y., Wright, Z., Steck, B., Luo, J., Frey, R., Coleman, A., Henderson-Brown, D., Berge, J., Caramori, M., Adeyi, O., Nachman, P., Safadi, S., Flanagan, S., Ma, S., Klett, S., Wolf, S., Harindhanavudhi, T., Rao, V., Mottl, A., Froment, A., Zeitler, E., Bream, P., Kelley, S., Rosengart, M., Elder, M., Palevsky, P., Murugan, R., Hall, D. E., Bender, F., Winters, J., Kellum, J. A., Gilliam, M., Tublin, M., Tan, R., Zhang, G., Sharma, K., Venkatachalam, M., Hendricks, A., Kermani, A., Torrealba, J., Vazquez, M., Wang, N., Cai, Q., Miller, R., Ma, S., Hedayati, S., Hoofnagle, A., Wangperawong, A., Berglund, A., Dighe, A. L., Young, B., Larson, B., Berry, B., Alpers, C., Limonte, C., Stutzke, C., Roberts, G., de Boer, I., Snyder, J., Phuong, J., Carson, J., Rezaei, K., Tuttle, K., Brown, K., Blank, K., Sarkisova, N., Jefferson, N., Mcclelland, R., Mooney, S., Nam, Y., Wilcox, A., Park, C., Dowd, F., Williams, K., Grewenow, S. M., Daniel, S., Shankland, S., Pamreddy, A., Ye, H., Montellano, R., Bansal, S., Pillai, A., Zhang, D., Park, H., Patel, J., Sambandam, K., Basit, M., Wen, N., Moe, O. W., Toto, R. D., Lee, S. C., Sharman, K., Caprioli, R. M., Fogo, A., Allen, J., Spraggins, J., Djambazova, K., de Caestecker, M., Dufresne, M., Farrow, M., Vijayan, A., Minor, B., Nwanne, G., Gaut, J., Conlon, K., Kaushal, M., Diettman, S. M., Victoria Castro, A. M., Moledina, D., Wilson, F. P., Moeckel, G., Cantley, L., Shaw, M., Kakade, V., Arora, T., Jain, S., Rauchman, M., Eadon, M. T. 2024; 15 (1): 433

    Abstract

    There is a need to define regions of gene activation or repression that control human kidney cells in states of health, injury, and repair to understand the molecular pathogenesis of kidney disease and design therapeutic strategies. Comprehensive integration of gene expression with epigenetic features that define regulatory elements remains a significant challenge. We measure dual single nucleus RNA expression and chromatin accessibility, DNA methylation, and H3K27ac, H3K4me1, H3K4me3, and H3K27me3 histone modifications to decipher the chromatin landscape and gene regulation of the kidney in reference and adaptive injury states. We establish a spatially-anchored epigenomic atlas to define the kidney's active, silent, and regulatory accessible chromatin regions across the genome. Using this atlas, we note distinct control of adaptive injury in different epithelial cell types. A proximal tubule cell transcription factor network of ELF3, KLF6, and KLF10 regulates the transition between health and injury, while in thick ascending limb cells this transition is regulated by NR2F1. Further, combined perturbation of ELF3, KLF6, and KLF10 distinguishes two adaptive proximal tubular cell subtypes, one of which manifested a repair trajectory after knockout. This atlas will serve as a foundation to facilitate targeted cell-specific therapeutics by reprogramming gene regulatory networks.

    View details for DOI 10.1038/s41467-023-44467-6

    View details for Web of Science ID 001141040600003

    View details for PubMedID 38199997

    View details for PubMedCentralID PMC10781985