A multidisciplinary team led by Associate Professor Mengyang Liu at the Center for Medical Physics and Biomedical Engineering, Medical University of Vienna, alongside Associate Professor Kristen Meiburger from Politecnico di Torino, have engineered a sophisticated OC-PAM platform specifically tailored for intricate 3D cancer models. By harnessing the complementary imaging capabilities of optical coherence microscopy (OCM) and photoacoustic microscopy (PAM), the researchers achieve label-free, volumetric visualization with remarkable spatial resolution and functional contrast. The integration of AI-driven analytics further empowers the system to perform comprehensive longitudinal tracking, viability assessments, and rare cell detection—capabilities that conventional imaging tools struggle to deliver without perturbation.

A key feat demonstrated by the OC-PAM system is its application to breast cancer organoids subjected to carboplatin chemotherapy. The use of OCM mode enabled high-resolution imaging of individual organoid structures over multiple time points, capturing dynamic volumetric changes associated with drug exposure. Automated algorithms tracked organoid morphology, revealing distinct growth trajectories that delineated responsive subpopulations from a resilient minority exhibiting regrowth—phenotypically consistent with drug-tolerant persister (DTP) cells. This ability to longitudinally monitor tumor heterogeneity and therapeutic resistance in a label-free setting represents a significant methodological leap.

Beyond morphological tracking, the research team introduced a radiomics-based framework that extracts high-dimensional quantitative features from OCM images. Paired with machine learning classifiers, this approach robustly discriminates viable from non-viable organoids without requiring invasive staining or fluorescence markers. The predictive accuracy achieved underscores the transformative potential of combining optical imaging with advanced computational analysis for continuous, non-destructive monitoring of treatment efficacy, a critical unmet need in preclinical oncology research.

The novel capabilities of the combined OC-PAM system extend to probing the cellular microenvironment within densely packed 3D spheroids, particularly the detection of rare cell populations. By co-culturing breast cancer cells with melanoma cells rich in melanin, the PAM component leveraged intrinsic optical absorption contrast to highlight these rare cell proxies amid the stromal milieu. Impressively, the system resolved individual melanoma cells even at extremely low concentrations, offering a sensitive platform for elucidating intratumoral heterogeneity and understanding the roles of minor subclones in cancer progression and drug resistance.

Central to the success of this approach is the synergy between OCM and PAM modalities. OCM provides label-free, volumetric imaging by exploiting the interference of backscattered near-infrared light, facilitating high axial and lateral resolution. Conversely, PAM transduces optically induced ultrasonic waves generated by transient thermoelastic expansion upon light absorption, revealing functional and molecular contrasts invisible to conventional microscopy. Their co-registration within the OC-PAM framework creates a comprehensive multimodal imaging channel that captures both structural and biochemical tumor attributes simultaneously.

The incorporation of AI-based analytics constitutes another pillar of this technological breakthrough. Leveraging convolutional neural networks and advanced radiomic feature extraction, the system automates organoid segmentation, classification, and viability scoring with minimal human intervention. This automated pipeline not only reduces observer bias but also accelerates data throughput, enhancing reproducibility and quantitative rigor for large-scale drug screening endeavors. It exemplifies the profound impact of integrating state-of-the-art optical hardware with computational intelligence in overcoming biological complexity.

Importantly, the ability to track drug-induced changes at the individual organoid level enhances the granularity of therapeutic assessment. Rather than summarizing responses across heterogeneous populations, this system reveals subtle variabilities within clonal populations, capturing early emergence of resistant phenotypes. Such insights are invaluable in informing adaptive therapy regimens and accelerating the identification of novel therapeutic targets specific to resistant cancer niches.

The researchers’ demonstration of non-invasive viability assessment further resonates with clinical aspirations for personalized medicine. By circumventing the need for destructive staining, the platform enables continuous, longitudinal studies on patient-derived organoids, potentially allowing on-demand evaluation of individualized drug responses. This capability aligns with the overarching goal of precision oncology to tailor treatment strategies based on dynamic tumor profiling rather than static, one-time biopsies.

Moreover, the success in detecting rare melanoma cells amidst breast cancer spheroids validates the platform’s application in modeling complex tumor-immune microenvironments and heterogeneous cell interactions. The exquisite sensitivity to minor cell subpopulations could catalyze studies on metastatic colonization, dormancy, and immune evasion—areas critical to unraveling cancer lethality mechanisms.

Taken together, this study establishes optical coherence photoacoustic microscopy combined with AI as a powerful imaging paradigm transcending traditional constraints. With its non-invasive, label-free, high-resolution, and functional imaging capabilities, along with robust computational analyses, OC-PAM is poised to revolutionize fundamental cancer biology research, drug development pipelines, and ultimately, precision oncology clinical workflows.

As translational researchers continue grappling with intratumor heterogeneity and therapeutic resistance, platforms like OC-PAM offer a glimpse into the future of cancer modeling—one where the dynamic interplay of cell populations can be visualized, quantified, and harnessed to devise more effective, personalized interventions. The fusion of cutting-edge optical technologies with AI analytics signals a new dawn in how complex cancer systems are interrogated and understood.

This technological breakthrough exemplifies how interdisciplinary convergence—melding optics, computational science, and cellular biology—can unlock new frontiers in biomedical imaging. The potential for scaling this approach, adapting it across diverse cancer types, and integrating functional assays positions OC-PAM as a cornerstone innovation in the fight against cancer.

In conclusion, the newly developed AI-enhanced optical coherence photoacoustic microscopy platform stands as a versatile, high-impact tool for the imaging and analysis of 3D cancer models. By overcoming the limitations of existing imaging methods, it enables unprecedented, label-free longitudinal studies of tumor organoids and spheroids. Such advancements promise to accelerate therapeutic discovery, illuminate mechanisms of drug resistance, and guide precision oncology with unparalleled resolution and depth.

Subject of Research: Optical coherence photoacoustic microscopy for imaging and AI-assisted analysis of 3D cancer organoids and spheroids.

Article Title: Optical coherence photoacoustic microscopy for 3D cancer model imaging with AI-assisted organoid analysis

Web References: DOI:10.1038/s41377-025-02177-2

Image Credits: Mengyang Liu et al.

Keywords: Optical coherence microscopy, photoacoustic microscopy, AI-assisted imaging, cancer organoids, spheroids, drug resistance, tumor heterogeneity, non-invasive imaging, longitudinal tracking, radiomics, drug-tolerant persister cells, melanoma cells detection

Leading this pioneering initiative is Xinlei (Sherry) Wang, the Jenkins Garrett professor of statistics and data science in UTA’s Department of Mathematics. Recently, Dr. Wang secured a prestigious four-year federal grant totaling $1.28 million to advance her project titled “Statistical and Deep Generative Modeling for Enhanced CyTOF Data Interpretation and Discovery.” This grant reflects the critical importance and potential impact of her interdisciplinary research that melds advanced statistics, deep learning, and biomedical science.

At the core of Wang’s research lies CyTOF (Cytometry by Time-Of-Flight), a cutting-edge technology that simultaneously quantifies dozens of proteins across thousands of individual cells. CyTOF generates exorbitantly complex datasets, capturing heterogeneity at the single-cell level—a necessary granularity for unraveling biological phenomena that are invisible in bulk analyses. However, this data’s complexity also presents a daunting challenge: how to transform this high-dimensional data into clear, interpretable insights usable by biomedical scientists who may lack computational expertise.

To tackle this challenge, Wang’s team employs a Bayesian statistical framework—an approach grounded in probability theory that quantifies uncertainty and integrates prior knowledge with observed data. Bayesian methods are particularly suited for biological data because they provide transparent and interpretable models, allowing researchers to infer biologically meaningful parameters such as protein expression differences between diseased and healthy cells. Wang’s group is developing unified statistical models that mechanistically characterize CyTOF data generation processes, thereby illuminating the hidden patterns and relationships embedded within.

The integration of artificial intelligence within this Bayesian framework dramatically enhances scalability and speed. Traditional computational techniques for single-cell data analysis can take several days to process millions of cells, impeding rapid discovery. By combining AI with Bayesian statistics, Wang’s models yield rigorous and reliable results within seconds. This remarkable acceleration doesn’t sacrifice interpretability; instead, it synergizes with the statistical rigor to enable both hypothesis generation and testing in real-time.

Crucially, Wang’s approach synthesizes data from single-cell transcriptomics and CyTOF protein profiling. Single-cell transcriptomics catalogs gene expression at an unprecedented scale, providing complementary information to the protein-level data generated by CyTOF. The joint analysis empowers researchers to capture a more comprehensive molecular portrait of cellular states and transitions. This integrated data fusion is pivotal for decoding complex biological circuits underlying diseases such as cancer, autoimmune disorders, and infectious diseases.

The resulting AI-driven toolkit can analyze millions of cells simultaneously, each characterized by 40 to 100 protein markers or tens of thousands of gene expression values. It excels at identifying distinct cell subtypes and comparing their molecular signatures across healthy and pathological conditions. By distinguishing subtle cellular differences, these models open new avenues for precision medicine, enabling tailored therapeutic interventions and improved prognostic assessments.

Wang’s team has already garnered significant recognition for their innovative contributions. Kevin Wang, a recent doctoral graduate mentored by Dr. Wang, received the Best PhD Poster Award at the 2025 Conference of Texas Statisticians for presenting their preliminary findings. This accolade underscores the growing impact of their work within the statistical community and its potential to transform biomedical research.

Further emphasizing their commitment to translational impact, the group recently published a study in the journal Nature Communications introducing BIT (Bayesian Identification of Transcriptional Regulators from Epigenomics-Based Query Regions Sets). BIT enhances the precision of identifying gene regulatory mechanisms by leveraging epigenomic data, a testament to the team’s expertise in bridging statistical modeling with cutting-edge genomic technologies.

The collaborative nature of Wang’s research extends beyond UTA. Key members include Li Wang, an associate professor of mathematics; Yike Shen, an assistant professor of earth and environmental sciences; and researchers at UT Southwestern such as Yuqiu Yang and Andy Xiao. Together, they form a multidisciplinary consortium advancing the frontiers of AI-driven biomedical data interpretation.

A critical innovation Wang highlights is the creation of user-friendly, open-source software packages that encapsulate the team’s complex algorithms while remaining accessible to end users. This democratization of technology ensures that researchers without extensive computational backgrounds can harness powerful AI tools on standard laptops. Existing methods often falter when confronted with big biological datasets, but Wang’s framework integrates statistical rigor, uncertainty quantification, and scalability to overcome these limitations seamlessly.

Dr. Wang aptly observes that although AI is potent, it is often a “black box” where decision-making processes are obscured. By embedding AI within transparent Bayesian models, her research restores interpretability, enabling users to understand the biological significance of algorithmic outputs and fostering trust in AI-driven discoveries.

The implications of this work are profound. As biomedical datasets continue to expand exponentially in both size and complexity, sophisticated analytical frameworks capable of delivering fast, interpretable, and scalable insights will be indispensable. Wang’s research not only addresses this need but sets a benchmark for integrating statistical theory, machine learning, and biological domain knowledge into a cohesive, practical system.

As the University of Texas at Arlington celebrates its 130th anniversary in 2025, this project exemplifies the institution’s growing stature as a Carnegie R-1 research university and its commitment to producing innovative solutions that impact health and society. With over 42,700 students and a significant economic influence in the Dallas-Fort Worth metroplex, UTA continues to foster groundbreaking research that pushes the boundaries of knowledge and technology.

In conclusion, Xinlei (Sherry) Wang’s research embodies the transformative potential at the nexus of AI, statistics, and biomedical science. By harnessing Bayesian methodologies and deep generative models to interpret CyTOF and transcriptomic data, her team is unveiling the cellular mysteries fundamental to disease processes. Their work offers a pathway to accelerated, interpretable, and large-scale biological discovery, with far-reaching implications for diagnosis, treatment, and prevention in medicine.

Subject of Research: Artificial intelligence application in Bayesian statistical modeling for single-cell CyTOF and transcriptomic data analysis.

Article Title: Statistical and Deep Generative Modeling for Enhanced CyTOF Data Interpretation and Discovery.

News Publication Date: Not explicitly stated (context suggests early 2025).

Web References: http://dx.doi.org/10.1038/s41467-025-60269-4

References:
Wang et al. (2025). “Bayesian Identification of Transcriptional Regulators from Epigenomics-Based Query Regions Sets” in Nature Communications.

Image Credits: UT Arlington

Keywords: Artificial intelligence, Bayesian statistics, CyTOF, single-cell analysis, deep generative modeling, transcriptomics, bioinformatics, statistical modeling, biomedical data interpretation, uncertainty quantification, scalable AI, cancer research.