Convolutional Neural Networks for the Evaluation of Chronic and Inflammatory Lesions in Kidney Transplant Biopsies

Hermsen, Meyke; Ciompi, Francesco; Adefidipe, Adeyemi; Denic, Aleksandar; Dendooven, Amélie; Smith, Byron H.; Midden, Dominique van; Bräsen, Jan Hinrich; Kers, Jesper; Stegall, Mark D.; Bándi, Péter; Nguyen, Tri Q.; Świderska-Chadaj, Żaneta; Smeets, Bart; Hilbrands, Luuk B.; Laak, J.A.W.M. van der

doi:10.1016/j.ajpath.2022.06.009

Cited by 21 publications

(16 citation statements)

References 48 publications

(47 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some studies previously described the potential of morphometric analysis of histology 19 , 27 – 32 . Although very specific, these scoring systems were applicable only in specific use cases and particular pathological alterations.…”

Section: Discussionmentioning

confidence: 99%

Next-Generation Morphometry for pathomics-data mining in histopathology

et al. 2023

View full text Add to dashboard Cite

Pathology diagnostics relies on the assessment of morphology by trained experts, which remains subjective and qualitative. Here we developed a framework for large-scale histomorphometry (FLASH) performing deep learning-based semantic segmentation and subsequent large-scale extraction of interpretable, quantitative, morphometric features in non-tumour kidney histology. We use two internal and three external, multi-centre cohorts to analyse over 1000 kidney biopsies and nephrectomies. By associating morphometric features with clinical parameters, we confirm previous concepts and reveal unexpected relations. We show that the extracted features are independent predictors of long-term clinical outcomes in IgA-nephropathy. We introduce single-structure morphometric analysis by applying techniques from single-cell transcriptomics, identifying distinct glomerular populations and morphometric phenotypes along a trajectory of disease progression. Our study provides a concept for Next-generation Morphometry (NGM), enabling comprehensive quantitative pathology data mining, i.e., pathomics.

show abstract

Section: Discussionmentioning

confidence: 99%

Next-Generation Morphometry for pathomics-data mining in histopathology

et al. 2023

View full text Add to dashboard Cite

show abstract

“…Some studies previously described the potential of morphometric analysis of histology 19,[27][28][29][30][31][32] . Although very specific, these scoring systems were applicable only in specific use cases and particular pathological alterations.…”

Section: Discussionmentioning

confidence: 99%

Next-Generation Morphometry for pathomics-data mining in histopathology

Hölscher

Bouteldja

Joodaki

et al. 2022

Preprint

View full text Add to dashboard Cite

Pathology diagnostics relies on the assessment of morphological features by trained experts, which remains subjective and qualitative. Modern image analysis techniques, particularly deep learning, provide a possible solution, sometimes exceeding human capabilities, e.g., mutation prediction directly from histology. 49 However, categorical model outputs are of limited use for further downstream analyses and limited interpretability. Here we developed a framework for large-scale histomorphometry (FLASH) which performs semantic segmentation and subsequent large-scale extraction of interpretable morphometric features. Two internal and three external, multi-centre cohorts of kidney biopsies were used to generate 40 million data points. Association with clinical data confirmed previous concepts, e.g., the importance of tubular atrophy for kidney function decline, and revealed unexpected findings, such as glomerular tuft hypertrophy in biopsies from patients with vs. without nephrotic range proteinuria. Single-structure analysis identified distinct glomerular populations and morphometric phenotypes along a trajectory of disease progression and features were independently associated with long-term clinical outcomes in IgA nephropathy. These data provide the concept for Next-generation Morphometry (NGM), opening new possibilities for comprehensive quantitative pathology data mining, i.e., pathomics, enabling augmented research and diagnostics.

show abstract

Section: Introductionmentioning

confidence: 99%

“…To date, several teams have successfully created deep learning networks (DLNs) that segment renal biopsies into their respective objects such as glomeruli, tubules, and vessels using conventional stains. [6][7][8][9] Other groups have focused on prediction of diagnosis. 10 However, no previous studies have employed automated image analysis in CD45 (leukocyte common antigen) stained slides to assess interstitial inflammation.…”

mentioning

confidence: 99%

Automated scoring of total inflammation in renal allograft biopsies

Smith

Grande

Ryan

et al. 2022

Clinical Transplantation

Self Cite

View full text Add to dashboard Cite

Background Computer‐assisted scoring is gaining prominence in the evaluation of renal histology; however, much of the focus has been on identifying larger objects such as glomeruli. Total inflammation impacts graft outcome, and its quantification requires tools to identify objects at the cellular level or smaller. The goal of the current study was to use CD45 stained slides coupled with image analysis tools to quantify the amount of non‐glomerular inflammation within the cortex. Methods Sixty renal transplant whole slide images were used for digital image analysis. Multiple thresholding methods using pixel intensity and object size were used to identify inflammation in the cortex. Additionally, convolutional neural networks were used to separate glomeruli from other objects in the cortex. This combined measure of inflammation was then correlated with rescored Banff total inflammation classification and outcomes. Results Identification of glomeruli on biopsies had high fidelity (mean pixelwise dice coefficient of .858). Continuous total inflammation scores correlated well with Banff rescoring (maximum Pearson correlation .824). A separate set of thresholds resulted in a significant correlation with alloimmune graft loss. Conclusions Automated scoring of inflammation showed a high correlation with Banff scoring. Digital image analysis provides a powerful tool for analysis of renal pathology, not only because it is reproducible and can be automated, but also because it provides much more granular data for studies.

show abstract

Convolutional Neural Networks for the Evaluation of Chronic and Inflammatory Lesions in Kidney Transplant Biopsies

Cited by 21 publications

References 48 publications

Next-Generation Morphometry for pathomics-data mining in histopathology

Next-Generation Morphometry for pathomics-data mining in histopathology

Next-Generation Morphometry for pathomics-data mining in histopathology

Automated scoring of total inflammation in renal allograft biopsies

Contact Info

Product

Resources

About