Validation of a deep learning-based image analysis system to diagnose subclinical endometritis in dairy cows

Abstract

The assessment of polymorphonuclear leukocyte (PMN) proportions (%) of endometrial samples is the hallmark for subclinical endometritis (SCE) diagnosis. Yet, a non-biased, automated diagnostic method for assessing PMN% in endometrial cytology slides has not been validated so far. We aimed to validate a computer vision software based on deep machine learning to quantify the PMN% in endometrial cytology slides. Uterine cytobrush samples were collected from 116 postpartum Holstein cows. After sampling, each cytobrush was rolled onto three different slides. One slide was stained using Diff-Quick, while a second was stained using Naphthol (golden standard to stain PMN). One single observer evaluated the slides twice at different days under light microscopy. The last slide was stained with a fluorescent dye, and the PMN% were assessed twice by using a fluorescence microscope connected to a smartphone. Fluorescent images were analyzed via the Oculyze Monitoring Uterine Health (MUH) system, which uses a deep learning-based algorithm to identify PMN. Substantial intra-method repeatabilities (via Spearman correlation) were found for Diff-Quick, Naphthol, and Oculyze MUH (r = 0.67 to 0.76). The intra-method agreements (via Kappa value) at ≥1% PMN (κ = 0.44 to 0.47) were lower than at >5 (κ = 0.69 to 0.78) or >10% (κ = 0.67 to 0.85) PMN cut-offs. The inter-method repeatabilities (via Lin’s correlation) were also substantial, and values between Diff-Quick and Oculyze MUH, Naphthol and Diff-Quick, and Naphthol and Oculyze MUH were 0.68, 0.69, and 0.77, respectively. The agreements among evaluation methods at ≥1% PMN were weak (κ = 0.06 to 0.28), while it increased at >5 (κ = 0.48 to 0.81) or >10% (κ = 0.50 to 0.65) PMN cut-offs. To conclude, deep learning-based algorithms in endometrial cytology are reliable and useful for simplifying and reducing the diagnosis bias of SCE in dairy cows.

Comparison of Sysmex XN-V body fluid mode and deep-learning-based quantification with manual techniques for total nucleated cell count and differential count for equine bronchoalveolar lavage samples

“…In case of the 800-cell differential count the observed drawbacks were a slightly longer software processing time and less opportunities to choose only monolayer areas where the cells were not touching each other or overlapping. These results prove the potential of AI to improve poorly precise manual laboratory methods and are in line with a previous study of bovine uterine cytobrush samples [ 27 ].…”

“…These observations are similar to our study where the best correlations were seen in the most numerous cell populations. Intra-method repeatability was also substantial in the mentioned study [ 27 ]. When thinking of future for AI algorithms in general one of their main advantages is the ability to rapidly count and differentiate thousands of cells without human bias or assistance with better precision than the manual methods.…”

“…In this study, a concordance with ground truth data (a term referring to real world data used in training of an algorithm) was assessed for both methods and the deep-learning algorithm partially outperformed the manual detection by showing a concordance of 85% whereas the manual detection reached 66 to 86% [ 26 ]. In a study of bovine endometrial cytology, a moderate to substantial agreement was obtained between the manual count via light microscopy and the deep-learning technology using Oculyze Monitoring Uterine Health system for detection of PMN cells once the threshold was set to > 10% PMN cells [ 27 ].…”

Canine Cerebrospinal Fluid Analysis Using Two New Automated Techniques: The Sysmex XN-V Body Fluid Mode and an Artificial-Intelligence-Based Algorithm

“…The accuracy of an AI algorithm in the recognition of monocytoid cells was noted to be high in those two studies but still lower than for the cell populations of lymphocytes and neutrophils [ 26 , 27 ]. Regarding research on AI cell recognition in veterinary medicine, a study of equine bronchoalveolar lavage showed correlations of between r = 0.85 and 0.92 with a manual method for the identification of alveolar macrophages, lymphocytes, neutrophils and mast cells [ 28 ], and a study with bovine uterine cytobrush samples demonstrated adequate agreement between AI and the manual method for >5% and >10% PMN cell thresholds [ 38 ].…”

Oxidative Stress Induces Bovine Endometrial Epithelial Cell Damage through Mitochondria-Dependent Pathways

“…We deemed that severe endometritis (Se) was present if the proportion of PMNs was >25% [ 38 , 39 , 40 ] in UD and was accompanied by exfoliation or necrosis of endometrial epithelial cells. In contrast, we assessed that moderate endometritis (Moe) was present if the proportion of PMNs was 18–25% [ 41 , 42 , 43 ], endometrial epithelial cells were flattened, and the proportion of mononuclear cells was 5–10%. Mild endometritis (Mie) was defined if the proportion of PMNs was 2–5% [ 44 , 45 , 46 ] and mononuclear cells were 3–5% in lamina propria.…”