Francesco Cisternino

Francesco Cisternino

Auto-generated profile, based on publicly available information

Francesco Cisternino

Human Technopole, Milan, Italy

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
13 July 2024 in Nature Communications
Verified
Journal Article
Open access
PDF file
FCFrancesco Cisternino
Francesco Cisternino
SOSara Ometto
Sara Ometto
SCSoumick Chatterjee
Soumick Chatterjee
EGEdoardo Giacopuzzi
Edoardo Giacopuzzi
ALAdam P. Levine
Adam P. Levine
CGCraig A. Glastonbury
Craig A. Glastonbury
As vast histological archives are digitised, there is a pressing need to be able to associate specific tissue substructures and incident pathology to disease outcomes without arduous annotation. Here, we learn self-supervised representations using a Vision Transformer, trained on 1.7 M histology images across 23 healthy tissues in 838 donors from the Genotype Tissue Expression consortium (GTEx). Using these representations, we can automatically segment tissues into their constituent tissue substructures and pathology proportions across thousands of whole slide images, outperforming other self-supervised methods (43% increase in silhouette score). Additionally, we can detect and quantify histological pathologies present, such as arterial calcification (AUROC = 0.93) and identify missing calcification diagnoses. Finally, to link gene expression to tissue morphology, we introduce RNAPath, a set of models trained on 23 tissue types that can predict and spatially localise individual RNA expression levels directly from H&E histology (mean genes significantly regressed = 5156, FDR 1%). We validate RNAPath spatial predictions with matched ground truth immunohistochemistry for several well characterised control genes, recapitulating their known spatial specificity. Together, these results demonstrate how self-supervised machine learning when applied to vast histological archives allows researchers to answer questions about tissue pathology, its spatial organisation and the interplay between morphological tissue variability and gene expression.

5 November 2024 in medRxiv
Verified
Preprint
Open access
PDF file
SOSara Ometto
Sara Ometto
Soumick ChatterjeeSoumick Chatterjee
Soumick Chatterjee
AVAndrea Mario Vergani
Andrea Mario Vergani
ALArianna Landini
Arianna Landini
SSSodbo Sharapov
Sodbo Sharapov
CGCraig A. Glastonbury
Craig A. Glastonbury
Biobank-scale imaging provides an unprecedented opportunity to characterise thousands of organ phenotypes, how they vary in populations and how they relate to disease outcomes. However, deriving specific phenotypes from imaging data, such as Magnetic Resonance Imaging (MRI), requires time-consuming expert annotation, limiting scalability, and does not exploit how information-dense such image acquisitions are. In this study, we developed a 3D diffusion autoencoder to derive latent phenotypes from temporally resolved cardiac MRI data of 71,021 UK Biobank participants. These phenotypes were reproducible, heritable ( h 2 = [4 - 18%]), and significantly associated with cardiometabolic traits and outcomes, including atrial fibrillation ( P = 8.5 × 10 −29 ) and myocardial infarction ( P = 3.7 × 10 −12 ). By using latent space manipulation techniques, we were able to learn, directly interpret and visualise what specific latent phenotypes are capturing in a given MRI. To establish the genetic basis of such traits, we performed a genome-wide association study, identifying 89 significant common variants ( P < 2.3 × 10 −9 ) across 42 loci, including seven novel loci. Extensive multi-trait colocalisation analyses (PP.H 4 > 0.8) linked variants across phenotypic scales, from intermediate cardiac traits to cardiac disease endpoints. For example, rs142556838 that falls in CCDC141 colocalises with a latent imaging phenotype and a diastolic blood pressure locus. Using single-cell RNA-sequencing data we map CCDC141 expression specifically to a population of ventricular cardiomyocytes. Finally, Polygenic Risk Scores (PRS) derived from latent phenotypes demonstrated predictive power for a range of cardiometabolic diseases and enabled us to successfully stratify the individuals into different risk groups. In conclusion, this study showcases the use of diffusion autoencoding methods as powerful tools for unsupervised phenotyping, genetic discovery and disease risk prediction using cardiac MRI data.

6 March 2025 in medRxiv
Verified
Preprint
Open access
PDF file
FCFrancesco Cisternino
Francesco Cisternino
YSYipei Song
Yipei Song
TPTim S. Peters
Tim S. Peters
RWRoderick Westerman
Roderick Westerman
GBGert J. de Borst
Gert J. de Borst
Sander van der LaanSander van der Laan
Sander van der Laan
Intraplaque haemorrhage (IPH) represents a critical feature of plaque vulnerability as it is robustly associated with adverse cardiovascular events, including stroke and myocardial infarction. How IPH drives plaque instability is unknown. However, its identification and quantification in atherosclerotic plaques is currently performed manually, with high inter-observer variability, limiting its accurate assessment in large cohorts. Leveraging the Athero-Express biobank, an ongoing study comprising a comprehensive dataset of histological, transcriptional, and clinical information from 2,595 carotid endarterectomy patients, we developed an attention-based additive multiple instance learning (MIL) framework to automate the detection and quantification of IPH across whole-slide images of nine distinct histological stains. We demonstrate that routinely available Haematoxylin and Eosin (H&E) staining outperformed all other plaque relevant Immunohistochemistry (IHC) stains tested (AUROC = 0.86), underscoring its utility in quantifying IPH. When combining stains through ensemble models, we see that H&E + CD68 (a macrophage marker) as well as H&E + Verhoeff-Van Gieson elastic fibers staining (EVG) leads to a substantial improvement (AUROC = 0.92). Using our model, we could derive IPH area from the MIL-derived patch-level attention scores, enabling not only classification but precise localisation and quantification of IPH area in each plaque, facilitating downstream analyses of its association and cellular composition with clinical outcomes. By doing so, we demonstrate that IPH presence and area are the most significant predictors of both preoperative symptom presentation and major adverse cardiovascular events (MACE), outperforming manual scoring methods. Automating IPH detection also allowed us to characterise IPH on a molecular level at scale. Pairing IPH measurements with single-cell transcriptomic analyses revealed key molecular pathways involved in IPH, including TNF-α signalling, extracellular matrix remodelling and the presence of foam cells. This study represents the largest effort in the cardiovascular field to integrate digital pathology, machine learning, and molecular data to predict and characterize IPH which leads to better understanding how it drives symptoms and MACE. Our model provides a scalable, interpretable, and reproducible method for plaque phenotyping, enabling the derivation of plaque phenotypes for predictive modelling of MACE outcomes.

29 June 2026 in Nature Communications
Verified
Journal Article
Open access
PDF file
SOSara Ometto
Sara Ometto
Soumick ChatterjeeSoumick Chatterjee
Soumick Chatterjee
AVAndrea Mario Vergani
Andrea Mario Vergani
ALArianna Landini
Arianna Landini
SSSodbo Sharapov
Sodbo Sharapov
CGCraig A. Glastonbury
Craig A. Glastonbury
Biobank-scale imaging provides an unprecedented opportunity to characterise how thousands of organ phenotypes vary in populations. However, deriving specific phenotypes from imaging data requires time-consuming expert annotation, limiting scalability. In this study, we develop a 3D diffusion autoencoder to derive latent phenotypes from temporally resolved cardiac MRI data of 71,017 UK Biobank participants. These phenotypes are reproducible, heritable (h2 = [4—18%]), and significantly associated with cardiometabolic traits. To establish the genetic basis of such traits, we perform a genome-wide association study, identifying 89 significant common variants (P < 2.3 × 10−9) across 42 loci, including seven novel loci. Extensive multi-trait colocalisation analyses (PP.H4 > 0.8) link variants across phenotypic scales, from intermediate cardiac traits to cardiac disease endpoints. In conclusion, this study showcases the use of diffusion autoencoding methods as powerful tools for unsupervised phenotyping, genetic discovery and disease risk prediction using cardiac MRI data.