Tissue response to ionizing radiation varies widely between different tissue, tumor types and radiation quality. The variation of radiation response may also vary between patients. Identification of the causes of this variability in radiation sensitivity could have important implications for cancer therapy. It is well known that the variance of specific energy imparted per cell increases with decreasing cell size. However, currently when studying the effects of morphological variations with the Monte Carlo method, properties such as energy absorption have been studied using a target geometry made up of single sized spheres, ellipsoids, or voxels to mimic the individual cells. While uniformly sized cells/nuclei as target geometry may provide useful information for many cases, any effects that a nuclei population with diverse volume sizes may have on the studied quantities are usually ignored due to lack of available data. In this presentation, I will discuss the use of deep learning models in automation of segmenting tumor regions in scanned images of histopathology slides, methods to derive cell/nucleus size distributions from these images and build digital 3D models for use in microdosimetry. In addition, microdosimetric spread due to variation in cell nuclei size will be discussed and compared with radiobiology studies.

Shirin Abbasinejad Enger is Canada Research Chair in Medical Physics and Associate Professor in the Gerald Bronfman Department of Oncology, Medical Physics Unit, McGill University where she conducts research in novel patient-specific brachytherapy delivery systems and Monte Carlo-based treatment planning software, deep learning-based image segmentation and dosimetry, detector development, radiobiology and microdosimetry. She received her PhD degree from Uppsala University, Sweden in 2009, and from 2009 to 2011 was a postdoctoral fellow at Université Laval. She joined McGill University in 2014.