Abstract Introduction: We aim to use clinical imaging data to initialize and constrain tumor-specific biophysical models of in situ cell proliferation, motility, and therapy response. Towards this goal, we have developed a 3D in vitro cell culture system that can be non-invasively interrogated by both microscopy and magnetic resonance imaging (MRI). The system supports long-term cell growth and allows for control over microenvironmental factors such as drug delivery and extracellular matrix. Cellular and subcellular level data obtained from confocal microscopy define the cell state and local environment and improve interpretation of the lower resolution, clinically relevant information provided by MRI. Here, we report system design and initial data collection. Apparatus Design: The system consists of the bioreactor body, a reservoir with a filter enabling CO2 equilibration in the incubator, and a peristaltic pump to drive media flow throughout the tubing and bioreactor. The bioreactor body has a cavity for cells embedded in extracellular matrix, a hollow-fiber for delivery of nutrients (FiberCell Systems, Frederick, MD), and is sealed using a gasket and lid with a confocal microscopy window. The removable lid makes the bioreactor reusable after sterilization. Data Collection: As a simplified model of 3D tumor growth, we selected the triple negative breast cancer cell line MDA-MB-231 embedded in 0.4% agarose. MDA-MB-231 cells were labeled with H2B-RFP and FUCCI-GFP to image cell nuclei and entry into the cell cycle, respectively, enabling measurements of cell number and cell division. Fluorescence microscopy images were collected on a Zeiss spinning disk confocal microscope using a 20x objective. Sixty 200 μm z -stacks using the green and red channels can be collected in approximately 90 minutes. The bioreactor is secured into a custom microscope mount using screws, ensuring it is placed in the same location every time. A reference point is marked on the first imaging day, to enable longitudinal alignment at subsequent imaging sessions. The bioreactor was also designed for use on a 9.4 T Agilent Technologies (Palo Alto, CA) MRI scanner equipped with a 38-mm quadrature RF coil, which is conventionally used for small animal studies. Diffusion-weighted (DW) images have been acquired using a standard pulsed gradient spin echo sequence with four b values (67, 200, 400, and 800 s/mm2) and gradients applied simultaneously along three orthogonal directions (x , y , and z ). The apparent diffusion coefficient obtained from the DW-MRI data provides an indirect measure of cell number. Images were acquired at a resolution of 100x100x1000 μm3. Conclusion: Overall, we have designed and fabricated a multi-modal imaging compatible bioreactor that supports long-term 3D cell growth under controlled conditions. We have successfully collected both MRI and confocal microscopy data on the bioreactor system at multiple time points. Currently, we are working towards co-registering these two imaging data types and extending the life of the bioreactor. Our novel multi-modal imaging compatible bioreactor is ideal for iterative biophysical tumor model refinement, providing a bridge to predictive models exclusively populated with non-invasive clinical imaging data. Citation Format: Abigail M. Searfoss, Matthew T. McKenna, Vito Quaranta, Thomas E. Yankeelov, Erin C. Rericha. A window into 3D culture: A multi-modal imaging compatible bioreactor for developing tumor growth models. [abstract]. In: Proceedings of the AACR Special Conference on Engineering and Physical Sciences in Oncology; 2016 Jun 25-28; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2017;77(2 Suppl):Abstract nr A22.