Organoid Models and AI Tools Drive Progress in Cancer Treatment Personalization

Organoids have emerged as next-generation tools for capturing the diversity of brain tumours in a clinically relevant context, offering a powerful preclinical and co-clinical platform with which to personalize neuro-oncology. These three-dimensional models recapitulate tumour heterogeneity and key pathophysiological processes in vitro, while remaining relatively time and cost effective to generate.

Machine learning algorithms are being developed that can extract insights from tumor imaging that only artificial intelligence can really see. The goal is to use that information to provide a treatment recommendation to the oncologist and patient. That kind of AI assist might be particularly useful in situations where there's clinical equipoise, when oncologists don't have clear indications which treatment will work better for an individual.

Tumour organoids effectively combine the benefits of basic cell lines and intricate in vivo models, enabling physiologically relevant ex vivo modelling of brain tumours. Brain tumour organoids include patient-derived tumour organoids, genetically engineered organoids and tumour assembloids, offering diverse characteristics and applicability. Primary patient-derived organoids can be derived from a large variety of brain tumours, positioning them as excellent ex vivo tumour avatars for functional readouts and precision medicine.

A team of European scientists have announced an artificial intelligence enhanced imaging platform that gives researchers a new way of studying cancer organoids and spheroids for non-invasive, label-free monitoring of tumour models over time. Researchers developed an AI-enhanced optical coherence photoacoustic microscopy system known as OC-PAM.

Through three carefully designed experiments, the team demonstrated that OC-PAM can perform longitudinal tracking of organoids, evaluate their response to chemotherapy, indicate individual organoid viability and identify proxies for drug tolerant persister cells. Crucially, all of these capabilities are achieved in a non-invasive and label-free way.

For longitudinal imaging, the researchers employed the optical coherence microscopy mode to examine breast cancer organoids following chemotherapy exposure through carboplatin administration. Using automated tracking of individual organoids alongside quantitative analysis of their average volumes, the team assessed how the models responded to treatment. Drug-treated organoids exhibited reduced growth rates. Notably, a small subset showed regrowth patterns consistent with drug-tolerant persister cells, rare cells believed to contribute to treatment resistance and relapse.

Beyond morphological changes, the study introduced a radiomics-based analysis of optical coherence microscopy data to evaluate organoid viability. By applying machine learning techniques, the researchers achieved high classification performance, showcasing the platform's potential for non-destructive monitoring of treatment response over time.

In a further experiment, the team explored the system's sensitivity to rare cell populations. Using OC-PAM, they imaged melanin-containing melanoma cells mixed with breast cancer cells within dense 3D spheroids. Even at very low concentrations, individual rare cells were successfully visualised.

Genetically engineered organoids allow researchers to gain mechanistic insight into human tumour biology within a controlled genetic background. Brain tumour assembloids facilitate the investigation of tumour-cell invasion and growth in interaction with healthy human brain structures.

If one treatment fails to work, as it inevitably does for some portion of patients, time is lost and patients must move on to the next line of treatment. If AI can provide clarity here, it could help oncology reach a new level of personalized medicine. With its non-invasive, label-free and longitudinal capabilities, the system could help to advance cancer biology, accelerate drug development and support more personalised approaches to oncology.