Abstract:
Automated analysis of digital pathology images when used in a tele-health setting can
have a transformative impact on underserved communities in the developing world.
However, the vast majority of existing image analysis algorithms are trained on slide
images acquired via expensive Whole-Slide Imaging (WSI) scanners. High scanner
cost is a key bottleneck preventing large-scale adoption of automated digital pathology
solutions in the developing countries. In this work, we investigate the possibility of
performing automated image analysis and quantification using images captured from
the eyepiece of a microscope via a smartphone. There are some challenges associated
with using a Microscope plus smartphone setup (e.g. non-uniform intensity and
illumination across different images). Will these factor have impact on the
performance of the machine learning algorithms? The mitosis detection application is
considered as a use case, Faster-RCNN is implemented for Mitosis detection and
Segmentation. Results are compared with a mitosis dataset of comparable size to
obtain an estimate of how much (if any) performance degradation results by employing
microscope images.
