Assessment of glomerular morphological patterns by deep learning algorithms

Abstract

Background Compilation of different morphological lesion signatures is characteristic of renal pathology. Previous studies have documented the potential value of artificial intelligence (AI) in recognizing relatively clear-cut glomerular structures and patterns, such as segmental or global sclerosis or mesangial hypercellularity. This study aimed to test the capacity of deep learning algorithms to recognize complex glomerular structural changes that reflect common diagnostic dilemmas in nephropathology. Methods For this purpose, we defined nine classes of glomerular morphological patterns and trained twelve convolutional neuronal network (CNN) models on these. The two-step training process was done on a first dataset defined by an expert nephropathologist (12,253 images) and a second consensus dataset (11,142 images) defined by three experts in the field. Results The efficacy of CNN training was evaluated using another set with 180 consensus images, showing convincingly good classification results (kappa-values 0.838–0.938). Furthermore, we elucidated the image areas decisive for CNN-based decision making by class activation maps. Finally, we demonstrated that the algorithm could decipher glomerular disease patterns coinciding in a single glomerulus (e.g. necrosis along with mesangial and endocapillary hypercellularity). Conclusions In summary, our model, focusing on glomerular lesions detectable by conventional microscopy, is the first sui generis to deploy deep learning as a reliable and promising tool in recognition of even discrete and/or overlapping morphological changes. Our results provide a stimulus for ongoing projects that integrate further input levels next to morphology (such as immunohistochemistry, electron microscopy, and clinical information) to develop a novel tool applicable for routine diagnostic nephropathology.

Natural Language Processing in Diagnostic Texts from Nephropathology

“…In this context, we have also recently published a paper on the classification of glomerular changes in histological images by means of convolutional neural networks (CNNs). Based on a defined small number of change patterns, we were able to diagnose entities defined by only a small number of patterns [ 7 ]. For instance, on the basis of images of a patients glomeruli, amyloidosis and diabetic glomerulopathy are easy to predict [ 1 , 7 , 8 ].…”

“…Based on a defined small number of change patterns, we were able to diagnose entities defined by only a small number of patterns [ 7 ]. For instance, on the basis of images of a patients glomeruli, amyloidosis and diabetic glomerulopathy are easy to predict [ 1 , 7 , 8 ]. A diagnosis like lupus nephritis, which can show a plethora of patterns over time and space (within one biopsy), is in contrast not predicable solely based on one glomerular change pattern [ 1 , 7 , 8 ].…”

“…For instance, on the basis of images of a patients glomeruli, amyloidosis and diabetic glomerulopathy are easy to predict [ 1 , 7 , 8 ]. A diagnosis like lupus nephritis, which can show a plethora of patterns over time and space (within one biopsy), is in contrast not predicable solely based on one glomerular change pattern [ 1 , 7 , 8 ]. Demonstrating that only a part of kidney tissues (in our case glomeruli) is not enough to make a correct diagnosis is not surprising.…”

“…The problem is not so much the preparation of a compilation of final diagnoses and images, but rather the laborious generation of correct annotations. As an example, in our recent publication, three experts spent several weeks classifying individual images in order to generate a sufficiently large data set [ 7 ]. For a segmentation task, where every part of the images needs to be labelled, the effort is significantly higher.…”

“…This leads to the idea of using the diagnosis text for analysis, in contrast to our recent work on image data [ 7 ]. In a sense, this text analysis is a surrogate for non-existent image data and image analysis tools.…”