Explainable Biomarkers for Automated Glomerular and Patient-Level Disease Classification

Abstract

Background Pathologists use multiple microscopy modalities to assess renal biopsies. Besides usual diagnostic features, some changes are too subtle to be properly defined. Computational approaches have the potential to systematically quantitate subvisual clues, provide pathogenetic insight, and link to clinical outcomes. To this end, a proof of principle study is presented demonstrating that explainable biomarkers through machine learning can distinguish between glomerular disorders at the light microscopy level. Methods The proposed system employed image analysis techniques and extracted 233 explainable biomarkers related to color, morphology, and microstructural texture. Traditional machine learning was then used to classify minimal change disease (MCD), membranous nephropathy (MN), and thin-basement membrane nephropathy (TBMN) diseases on a glomerular and patient-level basis. Results The final model combined the Gini feature importance set and Linear Discriminant Analysis classifier. Six morphological (nuclei-to-glomerular tuft area, nuclei-to-glomerular area, glomerular tuft thickness > 10, glomerular tuft thickness > 3, total glomerular tuft thickness, and glomerular circularity) and four microstructural texture features (luminal contrast using wavelets, nuclei energy using wavelets, nuclei variance using color vector LBP, and glomerular correlation using GLCM) were together the best performing biomarkers. Accuracies of 76.86% and 86.67% were obtained for glomerular and patient-level classification, respectively. Conclusion Computational methods using explainable glomerular biomarkers have diagnostic value and are compatible with our existing knowledge of disease pathogenesis. Furthermore, this algorithm can be applied to clinical datasets for novel prognostic and mechanistic biomarker discovery.

Machine learning in renal pathology

“…The biomarker feature extraction and machine learning model proposed in Basso et al (5) was replicated in this work. Pre-processing steps are first employed to exclude small glomeruli, as well as color normalization to remove color variability using a modified version of Reinhard's method (5).…”

“…The biomarker feature extraction and machine learning model proposed in Basso et al (5) was replicated in this work. Pre-processing steps are first employed to exclude small glomeruli, as well as color normalization to remove color variability using a modified version of Reinhard's method (5). Three sub-glomerular structures were automatically segmented using a modified Naïve Bayes classifier: (1) luminal (space inside the Bowman's capsule and the capillary lumen), (2) glomerular tuft (the glomerular basement membrane (GBM) and mesangial matrix), and (3) nuclei (5).…”

“…Pre-processing steps are first employed to exclude small glomeruli, as well as color normalization to remove color variability using a modified version of Reinhard's method (5). Three sub-glomerular structures were automatically segmented using a modified Naïve Bayes classifier: (1) luminal (space inside the Bowman's capsule and the capillary lumen), (2) glomerular tuft (the glomerular basement membrane (GBM) and mesangial matrix), and (3) nuclei (5). A series of biomarkers are extracted from the sub-glomerular structures, which includes color, morphological, and microstructural texture features, forming a total of 233 biomarkers (5).…”

“…For this we develop and fine-tune three pre-trained deep learning models with data augmentation and Grad-CAM for better performance and interpretability. The other method is based on a recently published (5) biomarker feature extraction model (BFE) and a traditional machine learning classifier that uses 233 explainable features for glomerular and patient-level classification. Our study is a starting point in glomerular disease diagnosis by machine learning and encourages the field to undertake more and larger studies with regards to kidney disease diagnosis.…”

The fractal and textural analysis of glomeruli in obese and non-obese patients

“…Also, it is possible this technique could in the future be used to supplement fractal method in terms of explaining changes in fractal dimension and lacunarity. 28 , 29 Other methods such as Fourier Transform Infrared (FTIR) micro-spectroscopy can be applied for tissue analysis, but it also has limitations caused due to different tissue preparation methods (Zhodi et al 30 ).…”