Relationship between Liquid-Based Cytology Preservative Solutions and Artificial Intelligence: Liquid-Based Cytology Specimen Cell Detection Using YOLOv5 Deep Convolutional Neural Network

Abstract

<b><i>Introduction:</i></b> Deep learning is a subset of machine learning that has contributed to significant changes in feature extraction and image classification and is being actively researched and developed in the field of cytopathology. Liquid-based cytology (LBC) enables standardized cytological preparation and is also applied to artificial intelligence (AI) research, but cytological features differ depending on the LBC preservative solution types. In this study, the relationship between cell detection by AI and the type of preservative solution used was examined. <b><i>Methods:</i></b> The specimens were prepared from five preservative solutions of LBC and stained using the Papanicolaou method. The YOLOv5 deep convolutional neural network algorithm was used to create a deep learning model for each specimen, and a BRCPT model from five specimens was also created. Each model was compared to the specimen types used for detection. <b><i>Results:</i></b> Among the six models, a difference in the detection rate of approximately 25% was observed depending on the detected specimen, and within specimens, a difference in the detection rate of approximately 20% was observed depending on the model. The BRCPT model had little variation in the detection rate depending on the type of the detected specimen. <b><i>Conclusions:</i></b> The same cells were treated with different preservative solutions, the cytologic features were different, and AI clarified the difference in cytologic features depending on the type of solution. The type of preservative solution used for training and detection had an extreme influence on cell detection using AI. Although the accuracy of the deep learning model is important, it is necessary to understand that cell morphology differs depending on the type of preservative solution, which is a factor affecting the detection rate of AI.

Relationship between a deep learning model and liquid‐based cytological processing techniques

“…The detection rate of TP preparations for the TS models was higher than that of the SP preparations in all five cell lines but was slightly lower than that of the TP model. This is consistent with our previous finding that the preparations used for training and detection should follow the same processing techniques to obtain a high detection rate with a high degree of accuracy 12. Although it may be possible to use a specific model based on the type of fixative, it is almost impossible to use a tumour-specific model for diagnostic samples.…”

“…The detection rate (approximately 25%) varies depending on the type of LBC preservative solution used. 12 According to the reports, the TP and SP specimens have different morphologies. [6][7][8][9][10] Mathematical and statistical analyses have revealed that the cytoplasm and nuclear areas of the TP specimens were larger than those of the SP specimens.…”

“…These differences also affect cell detection using AI. The detection rate (approximately 25%) varies depending on the type of LBC preservative solution used 12 . According to the reports, the TP and SP specimens have different morphologies 6–10 .…”

“…11 Previous studies have reported the relationship between LBC preservative solutions and cell detection using AI. 12 During cytological sample preparation, the differences in the LBC preservative solutions used for training and detection decrease cell detection accuracy. In addition, the construction of a deep learning model for use with specimens prepared under a variety of conditions is required.…”

“…Deep learning‐based studies often utilise models that are examined under limited clinical settings and trained on relatively small datasets with little heterogeneity and, therefore, cannot be generalised to a large number of laboratory settings 11 . Previous studies have reported the relationship between LBC preservative solutions and cell detection using AI 12 . During cytological sample preparation, the differences in the LBC preservative solutions used for training and detection decrease cell detection accuracy.…”