Seminars 23/24
The TIA seminar is a regular event, organised by the Tissue Image Analytics Centre, that usually takes place on the 1st and 3rd Mondays of each month between 2pm and 3pm. We invite researchers and leaders in the area of computational pathology and associated areas to discuss their work and stimulate thought-provoking discussions.
Take a look at the details of TIA seminars during 2022-23 academic year, most of which are also available on our YouTube channel.
Please contact the TIA Seminar Organiser Dr Adam Shephard via email (adam.shephard@warwick.ac.uk) if you are interested in attending any of the seminars below.
Summer 2024
| Date | Speaker | Title |
|
22nd April 2024 |
Daan Geijs Radbound UMC |
|
| 29th April 2024 |
Ali Khajegili Mirabadi University of British Columbia |
GRASP: GRAph-Structured Pyramidal Whole Slide Image Representation |
|
13th May 2024 |
Sebastian Klein University Duisburg-Essen |
|
|
10th June 2024 |
Drew Williamson Emory University |
|
|
24th June 2024 |
Stefan Feuerriegel and Valentyn Melnychuk LMU Munich |
|
|
27th June 2024 |
Inti Zlobec University of Bern |
|
|
22nd July 2024 |
Sajid Javed Khalifa University |
CPLIP: Zero-Shot Learning for Histopathology with Comprehensive Vision-Language Alignment |
Spring 2024
| Date | Speaker | Title |
|
15th January 2024 |
Gabriele Campanella Icahn School of Medicine at Mount Sinai |
|
|
22nd January 2024 |
Asmaa Ibrahim University of Nottingham |
Practical Application of Artificial Intelligence Models in Mitosis Scoring |
|
5th February 2024 |
Georg Wölflein University of St Andrews |
A Good Feature Extractor Is All You Need for Weakly Supervised Learning in Histopathology |
|
19th February 2024 |
Mark Eastwood University of Warwick |
TIAViz: An open-source visualization tool in TIAToolbox
|
|
4th March 2024 |
Hakim Benkirane CentraleSupelec |
|
|
5th March 2024 |
Sabine Tejpar and Stefan Naulaerts KU Leuven |
|
|
25th March 2024 |
Lena Cords Helmholtz Munich |
|
|
8th April 2024 |
Muhammad Shaban Harvard Medical School |
MAPS: pathologist-level cell type annotation from tissue images through machine learning
|
|
15th April 2024 |
Hamid Tizhoosh Mayo Clinic |
Learning or Searching: Foundation Models and Information Retrieval in Digital Pathology
|
Autumn 2023
| Date | Speaker | Title |
|
2nd October 2023 |
Zhi Huang Stanford University |
A visual–language foundation model for pathology image analysis using medical Twitter
|
|
16th October 2023 |
Yuri Tolkach University Hospital Cologne |
Using artificial intelligence for diagnostic and prognostic algorithms in gastrointestinal cancers
|
|
6th November 2023 |
Anglin Dent University of Toronto |
|
|
27th November 2023 |
Pingjun Chen University of Texas |
|
|
4th December 2023 |
Jan Brosens University of Warwick |
Endometrial diagnostics for the prediction and prevention of miscarriage |
|
11th December 2023 |
Andrew Song Harvard Medical School |
AI-driven efficient patient prognosis based on 3D pathology samples |
Autumn 2023
Dr Zhi Huang
Stanford University, Stanford, USA (click here for a short bio)
2 Oct 2023
Title: A visual–language foundation model for pathology image analysis using medical Twitter
Abstract: The lack of annotated publicly available medical images is a major barrier for computational research and education innovations. At the same time, many de-identified images and much knowledge are shared by clinicians on public forums such as medical Twitter. Here we harness these crowd platforms to curate OpenPath, a large dataset of 208,414 pathology images paired with natural language descriptions. We demonstrate the value of this resource by developing pathology language–image pretraining (PLIP), a multimodal artificial intelligence with both image and text understanding, which is trained on OpenPath. PLIP achieves state-of-the-art performances for classifying new pathology images across four external datasets: for zero-shot classification, PLIP achieves F1 scores of 0.565–0.832 compared to F1 scores of 0.030–0.481 for previous contrastive language–image pretrained model. Training a simple supervised classifier on top of PLIP embeddings also achieves 2.5% improvement in F1 scores compared to using other supervised model embeddings. Moreover, PLIP enables users to retrieve similar cases by either image or natural language search, greatly facilitating knowledge sharing. Our approach demonstrates that publicly shared medical information is a tremendous resource that can be harnessed to develop medical artificial intelligence for enhancing diagnosis, knowledge sharing and education.
Watch the seminar by clicking here.
Dr Yuri Tolkach
University Hospital Cologne, Cologne, Germany (click here for a short bio)
16 Oct 2023
Title: Using artificial intelligence for diagnostic and prognostic algorithms in gastrointestinal cancers
Abstract: In my talk I will cover how deep learning can be used for creation of diagnostic and prognostic algorithms for two types of GI cancers: oesophageal cancer and colorectal cancer. I will present information from three of our recent studies that show that AI-based algorithms can be seen as powerful assisting tools that allow for more precise diagnosis, creation of new and revision of old morphology-based prognostic parameters, and for saving pathologists' time on routine task automatization.
Watch the seminar by clicking here.
Anglin Dent
University of Toronto, Toronto, Canada (click here for a short bio)
6 Nov 2023
Title: HAVOC: Small-scale histomic mapping of cancer biodiversity across large tissue distances using deep neural networks
Abstract: Intratumoral heterogeneity can wreak havoc on current precision medicine strategies because of challenges in sufficient sampling of geographically separated areas of biodiversity distributed across centimeter-scale tumor distances. To address this gap, our group has developed a deep learning pipeline that leverages histomorphologic fingerprints of tissue to create “Histomic Atlases of Variation Of Cancers” (HAVOC). In this presentation, I will discuss how HAVOC can define regional cancer boundaries with distinct biology, map biodiversity across multiple tissue sections and tumor types, and reveal key biological insights of regionally distinct tumor populations distributed within these tumors.
Watch the seminar by clicking here.
Dr Pingjun Chen
University of Texas, Houston, USA (click here for a short bio)
27 Nov 2023
Title: Pathomic features reveal immune and molecular evolution from lung preneoplasia to invasive adenocarcinoma
Abstract: Lung adenocarcinoma (ADC) is the most common histologic subtype of lung cancer, and atypical adenomatous hyperplasia is the only recognized preneoplasia to ADC, which may progress to adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) and eventually to invasive ADC. Although molecular evolution during early lung carcinogenesis has been explored in recent years, the progress has been significantly hindered, primarily due to insufficient materials from ADC precursors. In this talk, I will present our study employing state-of-the-art deep learning and artificial intelligence techniques to robustly segment and recognize cells on routinely used H&E histopathology images and extract nine biology-relevant pathomic features to decode lung preneoplasia evolution. Extracted pathomic features revealed a progressive increase of atypical epithelial cells from normal to AAH, AIS, MIA, and ADC, consistent with the results from tissue-consuming and expensive molecular/immune profiling. Furthermore, pathomics analysis manifested progressively increasing cellular intratumor heterogeneity along with the evolution from normal lung to invasive ADC. These findings demonstrated the feasibility and substantial potential of pathomics in studying lung cancer carcinogenesis directly from the low-cost routine H&E staining.
Watch the seminar by clicking here.
Prof Jan Brosens
University of Warwick, Coventry, UK (click here for a short bio)
4 Dec 2023 (10:30am)
Title: Endometrial diagnostics for the prediction and prevention of miscarriage
Abstract: Compared to most mammals, human pregnancy is unusual in that it involves chromosomally diverse embryos, cyclical shedding and regeneration of the uterine mucosa (endometrium), and intimate integration of fetal and maternal cells at the uteroplacental interface. Not surprisingly, pregnancy often falters in early gestation. Whether these losses result in clinical miscarriages depends on the origins and impacts of chromosomal errors on fetal development and the ability of the cycling endometrium to transition successfully into the decidual of pregnancy, which anchors and supports the placenta throughout gestation. Aneuploidy originating in oocytes during meiosis drives the age-related risk of miscarriage. By contrast, the frequency of endometrial cycles with an impaired decidual response could theoretically account for the stepwise increase in miscarriage rates with each pregnancy loss independently of maternal age. Based on modelling human embryo implantation in 3D endometrial asssembloids and single cell transcriptomics, we discovered that functionally distinct endometrial stromal cell states render the endometrium receptive to embryo implantation and control the transition into the decidua of pregnancy, respectively. Analysis of over 1,500 endometrial biopsies from non-pregnant women revealed that an imbalance in cell state-specific biomarkers, normalised for the day of the biopsy in the cycle, correlates with the age-independent risk of miscarriage, has a high recurrence rate in different cycles, and is associated with increased risk of a future miscarriage. Further, we demonstrate that the pathological endometrial state associated with miscarriage is caused by loss of tissue plasticity. Current investigations are focussed on determining if machine learning methods on histology whole slide images enhance, complement or improve the performance of molecular methods in predicting pregnancy loss prior to conception.
Dr Andrew Song
Harvard Medical School, Boston, USA (click here for a short bio)
11 Dec 2023
Title: AI-driven efficient patient prognosis based on 3D pathology samples
Abstract: Human tissue forms a microenvironment that is fundamentally three-dimensional (3D). However, the standard-of-care in pathologic diagnosis involves selecting a few two-dimensional (2D) sections for microscopic evaluation, risking sampling bias and misdiagnosis. Diverse methods for capturing 3D tissue morphologies have been developed, with minimal success in translation to clinical practice; manual and computational evaluations of such large 3D data have been impractical and/or unable to provide patient-level clinical insights. Here we present Modality-Agnostic Multiple instance learning for volumetric Block Analysis (MAMBA), a deep-learning-based platform for processing 3D tissue images and predicting patient outcomes. Archived prostate cancer specimens were imaged with open-top light-sheet microscopy or microcomputed tomography and the resulting 3D datasets were used to train risk-stratification networks based on 5-year biochemical recurrence outcomes via MAMBA. With the 3D block-based approach, MAMBA achieves superior prediction performance to 2D traditional single-slice-based prognostication, suggesting superior prognostication with 3D morphological features. Further analyses reveal that the incorporation of greater tissue volume improves prognostic performance and mitigates risk prediction variability from sampling bias, suggesting the value of capturing larger extents of heterogeneous 3D morphology. With the rapid growth and adoption of 3D spatial biology and pathology, MAMBA provides a general and efficient framework for 3D pathology for clinical decision support and can help to reveal novel 3D morphological biomarkers for prognosis and therapeutic response.
Watch the seminar by clicking here.
Spring 2024
Dr Gabriele Campanella
Icahn School of Medicine at Mount Sinai, New York, USA (click here for a short bio)
15 Jan 2024
Title: Pathology Foundation Models at Health System Scale
Abstract: Recent breakthroughs in self-supervised learning have enabled the use of large unlabeled datasets to train visual foundation models that can generalize to a variety of downstream tasks. While this training paradigm is well suited for the medical domain where annotations are scarce, large-scale pre-training in the medical domain, and in particular pathology, has not been extensively studied. Previous work in pathology has focused on the use of self-supervised learning applied to smaller curated datasets for both pre-training and evaluating downstream performance. In this project we aim to train the largest academic pathology foundation model using health system scale data and to evaluate downstream performance on large clinical pathology datasets.
Watch the seminar by clicking here.
Dr Asmaa Ibrahim
University of Nottingham, Nottingham, UK (click here for a short bio)
22 Jan 2024
Title: Practical Application of Artificial Intelligence Models in Mitosis Scoring
Abstract: The implementation of AI in clinical practice needs to be critically evaluated against the existing methods. We aimed at assessing the optimal method of using AI-based mitotic figure scoring in breast cancer (BC). Utilizing whole slide images from a large Nottingham BC cohort (discovery n=1715, validation n=859) and TCGA-BRCA external test set (n=757), automated mitosis detection was applied using three methods: mitotic count per tumour area (MCT), mitotic index (MI), and mitotic activity index (MAI). The study compares these AI-based metrics with the Nottingham grading system (NGS), Ki67 score, clinicopathological parameters, and patient outcomes. While all three AI methods (MCT, MI, and MAI) significantly correlated with clinicopathological characteristics and patient survival (p<0.001), MAI and MCT show positive correlations with the gold standard visual scoring in NGS (r=0.8 and r=0.7, respectively) and Ki67 score (r=0.69 and r=0.55, respectively). MAI emerges as the sole independent predictor of survival (p<0.05) in multivariate Cox regression analysis. The findings emphasize the need to consider the optimal AI-based scoring method for clinical applications, highlighting MAI as a reliable and reproducible approach for accurate mitotic figure quantification in BC.
Watch the seminar by clicking here.
Georg Wölflein
University of St Andrews, St Andrews, UK (click here for a short bio)
5 Feb 2024
Title: A Good Feature Extractor Is All You Need for Weakly Supervised Learning in Histopathology
Abstract: Deep learning is revolutionising pathology, offering novel opportunities in disease prognosis and personalised treatment. Historically, stain normalisation has been a crucial preprocessing step in computational pathology pipelines, and persists into the deep learning era. Yet, with the emergence of feature extractors trained using self-supervised learning (SSL) on diverse pathology datasets, we call this practice into question. In an empirical evaluation of publicly available feature extractors, we find that omitting stain normalisation and image augmentations does not compromise downstream performance, while incurring substantial savings in memory and compute. Further, we show that the top-performing feature extractors are remarkably robust to variations in stain and augmentations like rotation in their latent space. Contrary to previous patch-level benchmarking studies, our approach emphasises clinical relevance by focusing on slide-level prediction tasks in a weakly supervised setting with external validation cohorts. This work represents the most comprehensive robustness evaluation of public pathology SSL feature extractors to date, involving more than 6,000 training runs across nine tasks, five datasets, three downstream architectures, and various preprocessing setups. Our findings stand to streamline digital pathology workflows by minimising preprocessing needs and informing the selection of feature extractors.
Watch the seminar by clicking here.
Dr Mark Eastwood
University of Warwick, Coventry, UK (click here for a short bio)
19 Feb 2024
Title: TIAViz: An open-source visualization tool in TIAToolbox
Abstract: Digital pathology has gained significant traction in modern healthcare systems. This shift from optical microscopes to digital imagery brings with it the potential for improved diagnosis and efficiency via the integration of AI tools into the pathologists workflow. A critical aspect of this is visualization. Throughout the development of a machine learning (ML) model in digital pathology, it is crucial to have flexible, openly available tools to visualize models, from their outputs and predictions to the underlying annotations and images used to train or test the model. TIAViz is a Python-based visualization tool built into open-source Digital Pathology package TIAToolbox which allows flexible, interactive, fully zoomable overlay of a wide variety of information onto whole slide images, including graphs, heatmaps, segmentations, annotations and other WSIs. The UI is browser-based, allowing use either locally, on a remote machine, or on a server to provide publicly available demos.
Watch the seminar by clicking here.
Hakim Benkirane
CentraleSupelec, Gif-sur-Yvettes, France (click here for a short bio)
4 March 2024
Title: Multimodal CustOmics: A Unified and Interpretable Multi-Task Deep Learning Framework for Multimodal Integrative Data Analysis in Oncology
Abstract: Characterizing cancer poses a delicate challenge as it involves deciphering complex biological interactions within the tumor's microenvironment. Histology images and molecular profiling of tumors are often available in clinical trials and can be leveraged to understand these interactions. However, despite recent advances in representing multimodal data for weakly supervised tasks in the medical domain, numerous challenges persist in achieving a coherent and interpretable fusion of whole slide images and multi-omics data. Each modality operates at distinct biological levels, introducing substantial correlations both between and within data sources. In response to these challenges, we propose a deep-learning-based approach designed to represent multimodal data for precision medicine in a readily interpretable manner. While demonstrating superior performance compared to state-of-the-art methods across multiple test cases, our approach also provides robust results and extracts various scores characterizing the activity of each modality and their interactions at the pathway and gene levels. The strength of our method lies in its capacity to unravel pathway activation through multimodal relationships and extend enrichment analysis to spatial data for supervised tasks. We showcase the efficiency and robustness of its predictive capacity and interpretation scores through an extensive exploration of multiple TCGA datasets and validation cohorts, underscoring its value in advancing our understanding of cancer.
Watch the seminar by clicking here.
Prof Sabine Tejpar and Dr Stefan Naulaerts
KU Leuven, Leuven, Belgium
5th March 2024
Title: Mining epithelial and immune biology in colorectal cancer
Dr Lena Cords
Helmholtz Munich, Munich, Germany (click here for a short bio)
25 March 2024
Title: Cancer-associated fibroblast phenotypes are associated with patient outcome in non-small cell lung cancer
Abstract: Despite advances in treatment, lung cancer survival rates remain low. A better understanding of the cellular heterogeneity and interplay of cancer-associated fibroblasts (CAFs) within the tumor microenvironment will support the development of personalized therapies. We report a spatially resolved single-cell imaging mass cytometry (IMC) analysis of CAFs in a non-small cell lung cancer cohort of 1,070 patients. We identify four prognostic patient groups based on 11 CAF phenotypes with distinct spatial distributions and show that CAFs are independent prognostic factors for patient survival. The presence of tumor-like CAFs is strongly correlated with poor prognosis. In contrast, inflammatory CAFs and interferon-response CAFs are associated with inflamed tumor microenvironments and higher patient survival. High density of matrix CAFs is correlated with low immune infiltration and is negatively correlated with patient survival. In summary, our data identify phenotypic and spatial features of CAFs that are associated with patient outcome in NSCLC.
Watch the seminar by clicking here.
Dr Muhammad Shaban
Harvard Medical School, Boston, USA (click here for a short bio)
8th April 2024
Title: MAPS: pathologist-level cell type annotation from tissue images through machine learning
Abstract: Highly multiplexed protein imaging is emerging as a potent technique for analyzing protein distribution within cells and tissues in their native context. However, existing cell annotation methods utilizing high-plex spatial proteomics data are resource intensive and necessitate iterative expert input, thereby constraining their scalability and practicality for extensive datasets. We introduce MAPS (Machine learning for Analysis of Proteomics in Spatial biology), a machine learning approach facilitating rapid and precise cell type identification with human-level accuracy from spatial proteomics data. Validated on multiple in-house and publicly available MIBI and CODEX datasets, MAPS outperforms current annotation techniques in terms of speed and accuracy, achieving pathologist-level precision even for typically challenging cell types, including tumor cells of immune origin. By democratizing rapidly deployable and scalable machine learning annotation, MAPS holds significant potential to expedite advances in tissue biology and disease comprehension.
Watch the seminar by clicking here.
Prof Hamid Tizhoosh
Mayo Clinic, Rochester, USA (click here for a short bio)
15th April 2024
Title: Learning or Searching: Foundation Models and Information Retrieval in Digital Pathology
Watch the seminar by clicking here.
Summer 2024
Daan Geijs
Radboud University Medical Center, Nijmegen, The Netherlands (click here for a short bio)
15th April 2024
22nd April 2024
Title: Dealing with the wave: Automating skin cancer assessment
Abstract: Faced with severe healthcare shortages and a basal cell carcinoma (BCC) incidence surpassing 40% of all cancer diagnoses in the Netherlands—an incidence expected to increase by an additional 39% by 2032—this end-of-PhD talk is dedicated to the research conducted to automate BCC detection. The goal is to automate this diagnostic processes and avoid further strain on healthcare systems.
Watch the seminar by clicking here.
Ali Khajegili Mirabadi
University of British Columbia, Vancouver, Canada (click here for a short bio)
29th April 2024
Title: GRASP: GRAph-Structured Pyramidal Whole Slide Image Representation
Abstract: Cancer subtyping is one of the most challenging tasks in digital pathology, where Multiple Instance Learning (MIL) by processing gigapixel whole slide images (WSIs) has been in the spotlight of recent research. However, MIL approaches do not take advantage of inter- and intra-magnification information contained in WSIs. In this work, we present GRASP, a novel graph-structured multi-magnification framework for processing WSIs in digital pathology. Our approach is designed to dynamically emulate the pathologist's behavior in handling WSIs and benefits from the hierarchical structure of WSIs. GRASP, which introduces a convergence-based node aggregation instead of traditional pooling mechanisms, outperforms state-of-the-art methods over two distinct cancer datasets by a margin of up to 10% balanced accuracy, while being 7 times smaller than the closest-performing state-of-the-art model in terms of the number of parameters. Our results show that GRASP is dynamic in finding and consulting with different magnifications for subtyping cancers and is reliable and stable across different hyperparameters. The model's behavior has been evaluated by two expert pathologists confirming the interpretability of the model's dynamic. We also provide a theoretical foundation, along with empirical evidence, for our work, explaining how GRASP interacts with different magnifications and nodes in the graph to make predictions. We believe that the strong characteristics yet simple structure of GRASP will encourage the development of interpretable, structure-based designs for WSI representation in digital pathology.
Watch the seminar by clicking here.
Dr Sebastian Klein
University Duisburg-Essen, Essen, Germany (click here for a short bio)
13th May 2024
Title: Predicting HPV association using deep learning and regular H&E stains allows granular stratification of oropharyngeal cancer patients
Abstract: Human Papilloma Virus (HPV)-associated oropharyngeal squamous cell cancer (OPSCC) represents an OPSCC subgroup with an overall good prognosis with a rising incidence in Western countries. Multiple lines of evidence suggest that HPV-associated tumors are not a homogeneous tumor entity, underlining the need for accurate prognostic biomarkers. In this seminar, we will discuss the clinical rational for a virtual imaging biomarker. We will visit the results of the published paper (Klein et al. 2023), and discuss the overall strategy of the computational approach.
Watch the seminar by clicking here.
Dr Drew Williamson
Emory University, Atlanta, USA (click here for a short bio)
10th June 2024
Title: Histologic Features: Everything Old is New Again
Abstract: Histologic features--essential characteristics that can be observed in a histology image--have a long history, one that dates back to the advent of histology. In this talk, we'll cover some of that history as well as methods for histologic feature extraction today. We'll also discuss some practical advice for researchers looking to extract those features from histology images and a future direction for this feature extraction based on prior knowledge of pathology.
Watch the seminar by clicking here.
Prof Stefan Feuerriegel and Valentyn Melnychuk
LMU Munich, Munich, Germany (click here for a short bio)
24th June 2024
Title: Causal machine learning for predicting treatment outcomes
Abstract: Causal machine learning (ML) offers flexible, data-driven methods for predicting treatment outcomes, with large potential for personalizing decision-making in medicine and management. Here, we explore recent advances in Causal ML and their relevance for translation in medicine and business decision-making, often motivated by practical considerations. (1) Many of the existing Causal ML methods are aimed at more standard settings but several, specialized settings in practice have been explored only recently (e.g., Causal ML for dosage combinations, Causal ML for time-series settings). (2) Causal ML often generates only point estimates while decision-making, especially in medicine, requires uncertainty estimates. (3) Causal ML rests on formal assumptions, which typically cannot be tested. Here, new methods such as in causal sensitivity analysis help improve the reliability in real-world settings.
Watch the seminar by clicking here.
Prof Inti Zlobec
University of Bern, Bern, Switzerland (click here for a short bio)
27th June 2024
Title: Does tumor budding really exist? How digital pathology helps to answer this question using 2D and 3D technologies
Abstract: The prognostic impact of tumor budding has now been established. But what do we really know about the nature of tumor buds? And do they even really “exist”? In this talk, we will follow tumor budding on a journey, which parallels the evolution of technologies for “tissue visualisation”, and look at the evidence from common staining approaches, to digital pathology & AI, spatial technologies in and 3D imaging, to help answer these questions.
Watch the seminar by clicking here.
Dr Sajid Javed
Khalifa University, Abu Dhabi, UAE (click here for a short bio)
22nd July 2024
Title: CPLIP: Zero-Shot Learning for Histopathology with Comprehensive Vision-Language Alignment
Abstract: This work proposes Comprehensive Pathology Language Image Pre-training (CPLIP), a new unsupervised technique designed to enhance the alignment of images and text in histopathology for tasks such as classification and segmentation. This methodology enriches vision-language models by leveraging extensive data without needing ground truth annotations. CPLIP involves constructing a pathology-specific dictionary, generating textual descriptions for images using language models, and retrieving relevant images for each text snippet via a pre-trained model. The model is then fine-tuned using a many-to-many contrastive learning method to align complex interrelated concepts across both modalities. Evaluated across multiple histopathology tasks, CPLIP shows notable improvements in zero-shot learning scenarios, outperforming existing methods in both interpretability and robustness and setting a higher benchmark for the application of vision-language models in the field. To encourage further research and replication, the code for CPLIP is available on GitHub at https://cplip.github.io/.
Watch the seminar by clicking here.