No abstract

Though deep learning has shown successful performance in classifying the label and severity stage of certain disease, most of them give few evidence on how to make prediction. Here, we propose to exploit the interpretability of deep learning application in medical diagnosis. Inspired by Koch's Postulates, a well-known strategy in medical research to identify the property of pathogen, we define a pathological descriptor that can be extracted from the activated neurons of a diabetic retinopathy detector. To visualize the symptom and feature encoded in this descriptor, we propose a GAN based method to synthesize pathological retinal image given the descriptor and a binary vessel segmentation. Besides, with this descriptor, we can arbitrarily manipulate the position and quantity of lesions. As verified by a panel of 5 licensed ophthalmologists, our synthesized images carry the symptoms that are directly related to diabetic retinopathy diagnosis. The panel survey also shows that our generated images is both qualitatively and quantitatively superior to existing methods.

No abstract

The traditional microscopic speckle interferometer has limited applications in engineering due to its small field of view. In this paper, we propose a large-field microscopic speckle interferometer which embeds two doublet lens groups in the improved Mach–Zehnder optical path structure to expand its field of view. At the same time, the new system can reduce the coherent noise of reflected light in the optical path. We use this new system to measure the dynamic displacement process of the entire surface of the microchips. The experimental results show that our improved measurement system can achieve large-field, real-time and high-precision dynamic measurement of micro-electromechanical systems (MEMS).

No abstract