A new, deep learning-based method for the analysis of autopsy kidney samples used to study sex differences in glomerular density and size in a forensic population.

Artificial intelligence (AI) is increasingly used in forensic anthropology and genetics to identify the victim and the cause of death. The large autopsy samples from persons with traumatic causes of death but without comorbidities also offer possibilities to analyze normal histology with AI. We propose a new deep learning-based method to rapidly count glomerular number and measure glomerular density (GD) and volume in post-mortem kidney samples obtained in a forensic population. We assessed whether this new method detects glomerular differences between men and women without known kidney disease. Autopsies performed between 2009 and 2015 were analyzed if subjects were aged ≥ 18 years and had no known kidney disease, diabetes mellitus, or hypertension. A large biopsy was taken from each kidney, stained with hematoxylin and eosin, and scanned. An in-house developed deep learning-based algorithm counted the glomerular density (GD), number, and size. Out of 1165 forensic autopsies, 86 met all inclusion criteria (54 men). Mean (± SD) age was 43.5 ± 14.6; 786 ± 277 glomeruli were analyzed per individual. There was no significant difference in GD between men and women (2.18 ± 0.49 vs. 2.30 ± 0.57 glomeruli/mm2, p = 0.71); glomerular diameter, area, and volume also did not differ. GD correlated inversely with age, kidney weight, and glomerular area. Glomerular area and volume increased significantly with age. In this study, there were no sex differences in glomerular density or size. Considering the size of the kidney samples, the use of the presented deep learning method can help to analyze large renal autopsy biopsies and opens perspectives for the histological study of other organs.

Reference values and sex differences in absolute and relative kidney size. A Swiss autopsy study.

BACKGROUND: Men have larger kidneys than women, but it is unclear whether gender remains an independent predictor of kidney size (expressed as weight or length) after correction for body size. We analysed autopsy data to assess whether relative renal length and weight (e.g. corrected for body weight, height or body surface area (BSA)) are also larger in men. Assuming that kidney size is associated with nephron number, opposite findings could partly explain why women are less prone to the development and progression of chronic kidney disease than men. METHODS: All forensic autopsies performed between 2009 and 2015 at the local university hospital of Geneva in individuals of European descent aged ≥18 years without a known history of diabetes and/or kidney disease were examined. Individuals with putrefied or severely injured bodies were excluded. Relative renal weight and length were respectively defined as renal weight divided by body weight or BSA and renal length divided by body height or BSA. RESULTS: A total of 635 autopsies (68.7% men) were included in the analysis. Left kidneys were on average 8 g heavier and 2 mm longer than right kidneys (both: p < 0.05). Absolute renal weight (165 ± 40 vs 122 ± 29 g) and length (12.0 ± 1.3 vs 11.4 ± 1.1 cm) were higher in men. Relative renal weight was also higher in men, but relative renal length was larger in women. In multivariable regression analysis, body height, body weight, the degree of blood congestion or depletion at autopsy and age were determinants of renal weight, whereas arterial hypertension and smoking were not. Percentile curves of renal weight and length according to sex and body height were constructed. CONCLUSION: Absolute and relative renal weights were both smaller in women. This is in line with recent studies stating that nephron numbers are also lower in women. Relative renal length was longer in women, suggesting that female kidneys have a more elongated shape. In comparison with older autopsy studies, renal weight appears to be stable over time.