Deep Neural Networks Offer Morphologic Classification and Diagnosis of Bacterial Vaginosis

Abstract

Background: Bacterial vaginosis (BV) is caused by the excessive and imbalanced growth of bacteria in vagina, affecting 30-50% of women in their lives. Gram stain followed by Nugent scoring based on bacterial morphotypes under the microscope (NS) has been considered the golden standard for BV diagnosis, which is often labor-intensive, time-consuming, and variable results from person to person. Methods: We developed and optimized a convolutional neural networks (CNN) model, and evaluated its ability to automatically identify and classify three categories of Nugent scores from microscope images. The CNN model was first established with a panel of microscopic images with Nugent scores determined by experts. The model was trained by minimizing the cross entropy loss function and optimized by using a momentum optimizer. The separate test sets of images collected from three hospitals were evaluated by the CNN models. Results: The CNN model consisted of 25 convolutional layers, 2 pooling layers, and a fully connected layer. The model obtained 82.4% sensitivity and 96.6% specificity on the 5,815 validation images when considered altered vaginal flora and BV as the positive samples, which was better than the top-level technologists and obstetricians in China. The ability of generalization for our model was strong that it obtained 75.1% accuracy of three categories of Nugent scores on the independent test set of 1082 images, which was 6.6% higher than the average of three technologists, who are with a bachelor degree in medicine and eligible making diagnostic decisions. When three technologists ran one specimen in triplicate, the precision of three categories of Nugent scores was 54.0%. 103 samples diagnosed by two technologists at different days showed repeatability of 90.3%. Conclusion: The CNN model over-performed human healthcare practitioners on accuracy and stability for three categories of Nugent scores diagnosis. The deep learning model may offer translational applications in automating diagnosis of bacterial vaginosis with proper supporting hardware.

Bacterial Vaginosis: What Do We Currently Know?

“…This model outperformed human healthcare professionals in terms of accuracy and stability for three diagnostic categories of Nugent scores. The deep learning model may offer translational applications in automating the diagnosis of BV with appropriate supporting hardware ( Wang et al., 2020 ).…”

Contribution of Lactobacillus iners to Vaginal Health and Diseases: A Systematic Review

“…By transmission electron microscopy, Kim et al revealed that the peptidoglycan (PG) layer in the cell wall of L. iners was thin enough to give an apparent Gram-negative morphology (Kim et al, 2020). This morphological characteristic and Gram-staining property of L. iners are clinically very important to consider, as Nugent scoring, which is based on the Gram-staining of vaginal smears, remains a common diagnostic tool in the assessment of vaginal health (Wang et al, 2021). The Gram-negative property of L. iners masks the fact that it is a Lactobacillus species and this may lead to the misdiagnosis of BV, which is a condition characterized by the depletion of Lactobacillus species in Gram-stained vaginal smears under microscopy (Vaneechoutte, 2017).…”

Aerobic Vaginitis Diagnosis Criteria Combining Gram Stain with Clinical Features: An Establishment and Prospective Validation Study

“…Applying this Gram-stain-based diagnostic criteria, the above-mentioned three common types of vaginal infection can be diagnosed on the same Gram-stained smear simultaneously, which improves the diagnostic efficiency and has good feasibility and generalizability. Recently, AI diagnosis of infectious diseases based on Gram stain has become an emerging interdisciplinary technology [ 25 , 38 ]. The proposal and application of these diagnostic criteria will lay a foundation for developing AV artificial intelligence diagnostic models.…”

“…Gram-stained smears can be magnified 1000 times to identify bacteria under ordinary optical microscopes and can be easily stored for a long time [ 24 ]. Moreover, diagnostic methods based on Gram staining make AI diagnosis possible [ 25 ]. Whether Gram-stained smears can follow the original wet-mount scoring system to diagnose AV remains to be elucidated.…”