Computational segmentation and classification of diabetic glomerulosclerosis

We developed a digital pipeline to classify renal biopsies from patients with Diabetic Nephropathy (DN). We combined traditional image analysis with modern machine learning to efficiently capture important structures, minimize manual effort and supervision, and enforce biologic prior information onto our model. To computationally quantify glomerular structure despite its complexity, we simplified it to three components consisting of nuclei, capillary lumina and Bowman spaces; and Periodic Acid-Schiff positive structures. We detected glomerular boundaries and nuclei from whole slide images using convolutional neural networks, and the remaining glomerular structures using an unsupervised technique developed expressly for this purpose. We defined a set of digital features which quantify the structural progression of DN, and a recurrent network architecture which processes these features into a classification.

Agreement statistics for human annotators and computational techniques in classifying DN biopsy cases.
Agreement statistics for human annotators and computational techniques in classifying DN biopsy cases.┬áReported values include linear weighted Cohen’s kappa and upper and lower 95% confidence intervals (CI). GT is ground truth, RP is renal pathologist, and C is computer. T refers to classifications according to the Tervaert scheme; F refers to classifications according to the Fogo scheme (defined by Dr. Agnes Fogo, Vanderbilt University).┬áData description identifies if the experiment was performed using separate patients or separate sections as individual data.

Select References

B. Ginley, B. Lutnick, K. Y. Jen, A. Fogo, S. Jain, A. Rosenberg, V. Walavalkar, G. Wilding, J. E. Tomaszewski, R. Yacoub, G. M. Rossi, and P. Sarder, “Computational segmentation and classification of diabetic glomerulosclerosis,” Journal of the American Society of Nephrology, vol. 30, no. 10, pp. 1953-67, Oct. 2019.