Artificial intelligence and machine learning have the potential to greatly transform healthcare. We will begin with highlighting a few applications in radiology, oncology and ophthalmology. Although these techniques have shown remarkable performance for many tasks including medical image analysis, we will share some of the challenges that we have faced in developing robust and trustworthy algorithms including a lack of repeatability, explainability, generalizability, and the potential for bias. We will conclude with a some mitigating strategies and recommendations.

The remarkable advances of artificial intelligence (AI) technology are revolutionizing established approaches to the acquisition, interpretation, and analysis of biomedical imaging data. Development, validation, and continuous refinement of AI tools requires  easy access to large high-quality annotated datasets, which are both representative and diverse. The National Cancer Institute (NCI) Imaging Data Commons (IDC) hosts over 50 TB of diverse publicly available cancer image data spanning radiology and microscopy domains. By harmonizing all  data based on industry standards and colocalizing it with analysis and exploration resources, IDC aims to facilitate the development, validation, and clinical translation of AI tools and address the well-documented challenges of establishing reproducible and  transparent AI processing pipelines. Balanced use of established commercial products with open-source solutions, interconnected  by standard interfaces, provides value and performance, while preserving sufficient agility to address the evolving needs of the research community. Emphasis on the development of tools, use cases to demonstrate the utility of uniform data representation, and  cloud-based analysis aim to ease adoption and help define best practices. Integration with other data in the broader NCI Cancer Research Data Commons infrastructure opens opportunities for multiomics studies incorporating imaging data to further empower the research community to accelerate breakthroughs in cancer detection, diagnosis, and treatment. The presentation will discuss the recent developments in IDC, highlighting resources, demonstrations and examples that we hope can help you improve your everyday imaging research practices - both those that use public and internal datasets.

AI for healthcare has the potential to revolutionize how we practice medicine. However, to do this in a fair and trustworthy manner requires special attention to how AI models work and their potential biases. In this talk, I will cover the considerations for building AI systems that improve healthcare.

For the development of ethical machine learning (ML) for precision medicine, it is essential to understand how values play into the decision-making process of developers. We conducted five group design exercises with four developer participants each (N=20) who were asked to discuss and record their design considerations in a series of three hypothetical scenarios involving the design of a tool to predict progression to diabetes. In each group, the scenario was first presented as a research project, then as development of a clinical tool for a health care system, and finally as development of a clinical tool for their own health care system. Throughout, developers documented their process considerations using a virtual collaborative whiteboard platform. Our results suggest that developers more often considered client or user perspectives after changing the context of the scenario from research to a tool for a large healthcare setting. Furthermore, developers were more likely to express concerns arising from the patient perspective and societal and ethical issues such as protection of privacy after imagining themselves as patients in the health care system. Qualitative and quantitative data analysis also revealed that developers made reflective/reflexive statements more often in the third round of the design activity (44 times) than in the first (2) or second (6) rounds. These statements included statements on how the activity connected to their real-life work, what they could take away from the exercises and integrate into actual practice, and commentary on being patients within a health care system using AI. These findings suggest that ML developers can be encouraged to link the consequences of their actions to design choices by encouraging “empathy work” that directs them to take perspectives of specific stakeholder groups. This research could inform the creation of educational resources and exercises for developers to better align daily practices with stakeholder values and ethical ML design.

In this seminar I will be addressing the dire cancer survival outcomes in low- and middle-income countries (LMICs), with a particular focus on Sub-Saharan Africa. Cancer survival rates in Sub-Saharan Africa are alarmingly low. According to the World Health Organization, cancer deaths in LMICs account for approximately 70% of global cancer fatalities. In Nigeria, the five-year survival rate for breast cancer, one of the most common cancers, stands at a disheartening 10-30%, compared to over 80% in high-income countries. This stark disparity highlights the urgent need for sustained comprehensive cancer interventions in our region. 

Here, I will discuss the pivotal role in the cancer control sphere, of a new software, ONCOSEEK, capable of early detecting 11 types of Cancers! It’s particular emphasis on the Patient Perspective, which aligns with our ethos of need for holistic patient care. In addition I will discuss recent developments on collaborative effort with the Gevaert lab at Stanford University and the University of Pennsylvania.

Evaluating radiology reports is a challenging problem as factual correctness is extremely important due to the need for accurate medical communication about medical images. Existing automatic evaluation metrics either suffer from failing to consider factual correctness (e.g., BLEU and ROUGE) or are limited in their interpretability (e.g., F1CheXpert and F1RadGraph). In this paper, we introduce GREEN (Generative Radiology Report Evaluation and Error Notation), a radiology report generation metric that leverages the natural language understanding of language models to identify and explain clinically significant errors in candidate reports, both quantitatively and qualitatively. Compared to current metrics, GREEN offers: 1) a score aligned with expert preferences, 2) human interpretable explanations of clinically significant errors, enabling feedback loops with end-users, and 3) a lightweight open-source method that reaches the performance of commercial counterparts. We validate our GREEN metric by comparing it to GPT-4, as well as to error counts of 6 experts and preferences of 2 experts. Our method demonstrates not only higher correlation with expert error counts, but simultaneously higher alignment with expert preferences when compared to previous approaches.

Recent advancements in self-supervised learning (SSL), emerging as an effective approach for imaging foundation models, enable the effective pretraining of AI models across multiple domains without the need for labels. Despite the rapid advancements, their application in medical imaging remains challenging due to the subtle difference between cancer and normal tissue. To address this limitation, in this study, we propose an AI architecture ProViCNet that employs the vision transformer (ViT) based segmentation architecture with patch-level contrastive learning for better feature representation. We validated our model in prostate cancer detection tasks using three types of magnetic resonance imaging (MRI) across multiple centers. To evaluate the performance of feature representation in this model, we performed downstream tasks with respect to Gleason grade score and race prediction. Our model demonstrated significant performance improvements compared to the state-of-the-art segmentation architectures. This study proposes a novel approach to developing foundation models for prostate cancer imaging overcoming SSL limitations.

Segmentation and synthesis are two fundamental tasks in medical image computing. Segmentation refers to the delineation of the boundaries of a structure of interest in the image, such as an organ, a tumor, or a lesion. Synthesis refers to images created computationally from other data; common examples include cross-modality synthesis and image denoising. This talk will provide an overview of my lab’s recent work in these two broad algorithmic directions in the context of a wide range of medical imaging applications. These driving clinical problems include MR imaging of the brain, OCT imaging of the retina, ultrasound imaging of the placenta, and endoscopic imaging of the kidney. I will also illustrate many problem formulations where synthesis can be used to help segmentation, and vice versa.

Dr. Aerts’ upcoming talk delves into the transformative role of artificial intelligence, especially deep learning, in medical imaging, focusing on pushing the boundaries of cancer care. As technology advances, deep learning models have become increasingly adept at image recognition and interpretation, unlocking new potentials in radiology, oncology, and beyond. Dr. Aerts will highlight some of the most recent developments from his research group and collaborators. This illustrates how AI’s capabilities are reshaping the role of medical imaging along the cancer care pathway, from risk prediction and diagnosis to treatment selection and monitoring response. A central theme will be foundational models—AI systems that, once trained on broad unlabeled data sets, can be fine-tuned for various tasks across different medical domains. These models promise to improve accuracy, especially in limited data scenarios. Additionally, Dr. Aerts will discuss the significance of open-source, transparent, and reproducible deep learning frameworks. In a field as sensitive and crucial as healthcare, these qualities are essential for building trust, ensuring reliability, and enabling other researchers to verify and build upon other’s work. He will address how these frameworks foster collaborative research and innovation across institutions, making advanced tools accessible and pushing the field forward.

Cancer is caused by genetic (DNA) and epigenetic alterations and frequently arises as a clonal growth from a founder cell. The subclonal heterogeneity provides the basis for inter-metastatic heterogeneity which is of utmost clinical importance. New tumor sampling techniques and circulating tumor DNA methods may allow for more comprehensive evaluation of clonal composition. As both primary tumors and metastatic lesions are spatially and temporally heterogeneous they would require multiple biopsies to extract and analyze small portions of tumor tissue, which still doesn’t allow for a complete characterization of the tumor genomic landscape. Therefore, imaging has great potential for a comprehensive evaluation of the entire tumor burden in ovarian cancer as it is noninvasive and is already often repeated during treatment in routine practice, on the contrary of genomics or proteomics, which are still challenging to implement into clinical routine. While initial retrospective studies linking phenotype with genotype in ovarian cancer have shown high prognostic power they do not provide any spatial information as quantitative imaging features are generated and averaged over the entire tumor assuming that tumors are heterogeneous but well mixed. This approach ignores spatial heterogeneity readily apparent on imaging. Indeed recent genomics work has highlighted the presence of intratumor variation in gene mutation and expression. However little effort is any has been put into integrating imaging, histopathology and genomics and thus there is a clear need for well designed prospective studies focused on meaningful integration of phenotype and genotype rather than genomics in isolation.