Testicular cancer is the most common cancer in young men, and seminoma is its most frequent type1. Most men with seminoma are cured after surgery removing the testicle (orchiectomy), but about 1 in 5 will later experience a relapse. Unfortunately, doctors cannot reliably predict at diagnosis who will relapse. Currently, factors such as tumor size or the presence of tumor cells in blood vessels are not very accurate, which means that some men receive potentially harmful and unnecessary additional treatment while others are under-treated.
Our project will explore improved ways to predict which patients are at risk for relapse. We are using digital pathology, where ML-based image processing algorithms analyze microscope images of tumor tissue to find patterns that pathologists cannot see with the naked eye. Early work shows this approach can identify subtle features linked to relapse, but the results are often without biological explanation, something that limits translation of this knowledge.
To make these tools more useful and trustworthy, we will combine digital pathology with spatial omics with the goal of uncovering the hidden biology that drives drives prognostication. For example, if digital pathology highlights a in the tissue, spatial omics can reveal whether this area shows any signs of immune evasion, low oxygen, or aggressive growth.
By combining these approaches, our team hopes to create more accurate and biologically meaningful tools to guide care for seminoma patients and improve their quality of life.
Seminoma tumors affects young men often during their most productive years of life. While the majority are cured with surgery, about 20% will relapse and require additional therapy. Because current tools cannot accurately predict relapse, many patients face difficult choices: undergo unnecessary adjuvant treatment and risk long-term toxicities, or remain on surveillance with the anxiety of possible relapse. Both pathways carry significant emotional, physical, and financial burdens to the patient.
This project directly addresses this gap by developing new tools that combine AI and computational pathology with cutting-edge molecular profiling of tissue (spatial omics). This integration has the potential to deliver more accurate and biologically meaningful predictors of relapse at the time of diagnosis. For patients, this means improved information and more confidence in decision-making, leading to personalized treatment strategies that avoid both over-treatment and under-treatment.
Additionally, our novel proposed work is expected to unlock new biology and generate novel biological insights into seminoma progression, potentially leading to the discovery of novel therapeutic targets/biomarkers that could benefit future patients.
Importantly, this project is at its core a truly collaborative framework and includes strong training and mentorship components, promoting interdisciplinary training of clinician-scientists and AI experts.