AI-algorithm training and validation for identification of endometrial CD138+ cells in infertility-associated conditions; polycystic ovary syndrome (PCOS) and recurrent implantation failure (RIF)

Abstract

Background: Endometrial CD138+ plasma cells serve as a diagnostic biomarker for endometrial inflammation, and their elevated occurrence correlates positively with adverse pregnancy outcomes. Infertility-related conditions like polycystic ovary syndrome (PCOS) and recurrent implantation failure (RIF) are closely associated with systemic and local chronic inflammatory status, wherein endometrial CD138+ plasma cell accumulation could also contribute to endometrial pathology. Current methods for quantifying CD138+ cells typically involve laborious and time-consuming microscopic assessments of only a few random areas from a slide. These methods have limitations in accurately representing the entire slide and are susceptible to significant biases arising from intra- and interobserver variations. Implementing artificial intelligence (AI) for CD138+ cell identification could enhance the accuracy, reproducibility, and reliability of analysis.

Methods: Here, an AI algorithm was developed to identify CD138+ plasma cells within endometrial tissue. The AI model comprised two layers of convolutional neural networks (CNNs). CNN1 was trained to segment epithelium and stroma across 28,363 mm² (2.56 mm² of epithelium and 24.87 mm² of stroma), while CNN2 was trained to distinguish stromal cells based on CD138 staining, encompassing 7345 cells in the object layers (6942 CD138- cells and 403 CD138+ cells). The training and performance of the AI model were validated by three experienced pathologists. We collected 193 endometrial tissues from healthy controls (n = 73), women with PCOS (n = 91), and RIF patients (n = 29) and compared the CD138+ cell percentages based on cycle phases, ovulation status, and endometrial receptivity utilizing the AI model.

Results: The AI algorithm consistently and reliably distinguished CD138- and CD138+ cells, with total error rates of 6.32% and 3.23%, respectively. During the training validation, there was a complete agreement between the decisions made by the pathologists and the AI algorithm, while the performance validation demonstrated excellent accuracy between the AI and human evaluation methods (intraclass correlation; 0.76, 95% confidence intervals; 0.36-0.93, p = 0.002) and a positive correlation (Spearman's rank correlation coefficient: 0.79, p < 0.01). In the AI analysis, the AI model revealed higher CD138+ cell percentages in the proliferative phase (PE) endometrium compared to the secretory phase or anovulatory PCOS endometrium, irrespective of PCOS diagnosis. Interestingly, CD138+ percentages differed according to PCOS phenotype in the PE (p = 0.03). On the other hand, the receptivity status had no impact on the cell percentages in RIF samples.

Conclusion: Our findings emphasize the potential and accuracy of the AI algorithm in detecting endometrial CD138+ plasma cells, offering distinct advantages over manual inspection, such as rapid analysis of whole slide images, reduction of intra- and interobserver variations, sparing the valuable time of trained specialists, and consistent productivity. This supports the application of AI technology to help clinical decision-making, for example, in understanding endometrial cycle phase-related dynamics, as well as different reproductive disorders.

Keywords: Artificial intelligence; CD138; Chronic endometritis; Computational histology; Polycystic ovary syndrome; Recurrent implantation failure.

Fig. 1

The study subjects in the control, PCOS, and RIF population. A total of 164 endometrial biopsy samples were collected from 44 healthy controls, 61 women with PCOS, and 29 RIF patients. The subjects were categorized based on whether they provided a single or multiple samples. In the case of women with PCOS, the subjects were additionally classified based on their ovulatory status. HRT (hormone replacement treatment); P4 (progesterone).

Fig. 2

Schematic overview of the AI algorithm and examples of training and validation. (a) The structure of the convolutional neural networks (CNNs) model includes CNN1 for the regional layer and CNN2 for the object layer. (b)–(d) Examples of the AI algorithm training, which was conducted through manual annotation. Only the areas within the regions of interest (ROIs) outlined as solid black lines were considered in the training. (b) original image, (c) manually annotated cells. (d) The AI algorithm training result. CD138− cells were marked cyan, while CD138+ cells were marked dark blue. (e)–(g) The training validation of the AI algorithm, where the validation regions were marked as black dotted line. (e) Analyzed images by the AI algorithm (CD138− cells marked cyan, CD138+ cells marked dark blue), (f) the validation from validator 1 (CD138- cells marked green, CD138+ cells marked yellow), (g) the validation from validator 2 (CD138− cells marked pink, CD138+ cells marked purple). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Stromal CD138+ plasma cell percentages across PCOS status and cycle phases. Stromal CD138+ cell percentages for: (a) controls (12 PE, 15 ESE, 25 MSE, 19 LSE) and PCOS cases (21 PE, 14 ESE, 21 MSE, 16 LSE, 11 Anovulatory) and (b) different PCOS phenotypes (Phenotype A; n = 12 PE, n = 26 SE, n = 8 Anovulatory, Phenotype D; n = 9 PE, n = 16 SE, n = 2 Anovulatory). Phenotype C was excluded due to the small sample size. The box indicates the interquartile range, the middle line represents the median, and the whiskers show the min–max range. ***p < 0.001 compared to the PE samples. The statistical differences were calculated by the mixed-model ANOVA. PE (proliferative phase), ESE (early secretory phase), MSE (mid-secretory phase), LSE (late secretory phase), SE (secretory phase).

Fig. 4

Stromal CD138+ plasma cell percentages stratified by endometrial receptivity. The CD138+ cell percentage comparisons of RIF patient samples (n = 9 RIF pre-receptive, n = 10 RIF receptive, n = 10 RIF post-receptive). The box indicates the interquartile range, the middle line represents the median, and the whiskers show the min–max range. The Kruskal–Wallis test did not reveal any significant differences between the groups.