Developing a Machine Learning Algorithm for Identifying Abnormal Urothelial Cells: A Feasibility Study

Abstract

Introduction: Urine cytology plays an important role in diagnosing urothelial carcinoma (UC). However, urine cytology interpretation is subjective and difficult. Morphogo (ALAB, Boston, MA, USA), equipped with automatic acquisition and scanning, optical focusing, and automatic classification with convolutional neural network has been developed for bone marrow aspirate smear analysis of hematopoietic diseases. The goal of this preliminary study was to determine the feasibility of developing a machine learning algorithm on Morphogo for identifying abnormal urothelial cells in urine cytology slides. Methods: Thirty-seven achieved abnormal urine cytology slides from cases with the diagnosis of atypical urothelial cells and above (suspicions or positive for UC) were obtained from 1 hospital. A pathologist (J.R.) reviewed the slides and manually selected and annotated representative cells to feed into Morphogo with following categories: benign (urothelial cells, squamous cells, degenerated cells, and inflammatory cells), atypical cells, and suspicious cells. Initial validation of the algorithm was performed on a subset of the original 37 cases. Urine samples from additional 12 unknown cases with various histological diagnoses (6 cases of high-grade urothelial carcinoma (HGUC), 1 case of lowgrade urothelial carcinoma (LGUC), 1 case of prostate adenocarcinoma, 1 case of renal cell carcinoma, and 4 cases of nonneoplastic conditions) were collected from another hospital for initial blind testing. Results: A total of 1,910 benign and 1,978 abnormal (atypical and suspicious) cells from 37 slides were annotated for developing and training of the algorithm. This algorithm was validated on 27 slides that resulted in identification of at least 1 abnormal cell per slide, with a total of 200 abnormal cells, and an average of 7.4 cells per slide. Of the 12 unknown cases tested, the original cytology was positive for tumor cells in 2 HGUC samples. Morphogo was abnormal (atypical or suspicious) for 6 samples from patients with UC, including one with LGUC and one with prostate adenocarcinoma. Conclusion: Morphogo machine learning algorithm is capable of identifying abnormal urothelial cells. Further validation studies with a larger number of urine samples will be needed to determine if it can be used to assist the cytological diagnosis of UC.

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

“…Object detection algorithms based on deep learning have advanced rapidly in recent years, and cytological research on AI using deep learning has been actively conducted with regard to gynecology [19][20][21], respiratory samples [22,23], body cavity fluid [24], urine samples [25,26], and thyroid tissues [27]. Most of these studies attempted to detect atypical cells and used various deep learning algorithms to improve the detection accuracy, but in many studies, programs were trained on relatively small datasets with little heterogeneity [1].…”

Recent Application of Artificial Intelligence in Non-Gynecological Cancer Cytopathology: A Systematic Review

“…The characteristics of the AI models in non-GYN cancer cytology are summarized in Table 1 . All 26 studies were published from March 2010 to March 2021 and conducted universally, including nine models in India, seven models in the USA, four models in Japan, three models each in China, two models in Greece, and one model in the UK, as shown in Figure 2 [ 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 , 38 , 39 , 40 , 41 , 42 , 43 , 44 , 45 , 46 , 47 , 48 , 49 ]. Further, we classified the algorithmic features of the models with respect to eight target organs: thyroid ( n = 11, 39%); urinary bladder ( n = 6, 21%); lung ( n = 4, 14%); breast ( n = 2, 7%); pleural effusion ( n = 2, 7%); ovary ( n = 1, 4%); pancreas ( n = 1, 4%); and prostate ( n = 1, 4%) ( Figure 3 A).…”

“…Six studies were included in the cytology classification of urine, and the dataset used ranged from 49 image patches to 2405 WSIs [ 42 , 43 , 44 , 45 , 46 , 47 ]. The AI models were trained to classify the urine samples of patients into three to four histological types, such as benign, low-grade, and high-grade urothelial carcinomas, or to classify the cell clusters into specific cell types, such as benign, atypical, and malignant urothelial cells, squamous cells, crystals, erythrocytes, leukocytes, blurry images, debris, degenerated cells, and inflammatory cells ( Table 1 ).…”

A Calibrated Multiexit Neural Network for Detecting Urothelial Cancer Cells

“…We refer the interested reader to specific surveys [12,[21][22][23][24] for an overview of CNN applications in medical histology and cytology. In the context of urinary cytopathology, a number of works [4,25,26] have shown promising preliminary results in automatically detecting carcinomas from urinary imaging. All of them, however, have focused on using out-of-the-box CNNs that are fine-tuned to the medical domain, while in this paper we design a more sophisticated architecture based on the two objectives described in Introduction.…”

“…In particular, today the examination of UT specimens in the cytopathology laboratory is typically used to screen for urothelial neoplasms in two populations: patients with new-onset and patients with a history of urothelial neoplasia, as urothelial carcinomas have high recurrence rates [ 3 ]. UC samples constitute a significant percentage of daily nongynecologic cases in any cytopathology laboratory and are one of the most difficult specimens that pathologists encounter [ 4 ].…”