Hybrid AI-assistive diagnostic model permits rapid TBS classification of cervical liquid-based thin-layer cell smears

Zhu, Xiaohui; Li, Xiaoming; Ong, Kokhaur; Zhang, Wenli; Li, Wencai; Li, Longjie; Young, David M.; Su, Yongjian; Shang, Bin; Peng, Linggan; Xiong, Wei; Liu, Yunke; Liao, Wenting; Xu, Jingjing; Wang, Feifei; Liao, Qing; Li, Shengnan; Liao, Minmin; Yu, Li; Rao, Linshang; Lin, Jinquan; Shi, Jing; You, Zejun; Zhong, Weiguo; Xin-rong, Liang; Hao, Han; Zhang, Yan; Tang, Na; Hu, Aixia; Gao, Hong-Yi; Cheng, Zhiqiang; Li, Liang; Yu, Weimiao; Ding, Yanqing

doi:10.1038/s41467-021-23913-3

Cited by 47 publications

(26 citation statements)

References 57 publications

(54 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Increasingly, studies are confirming and supporting AI technology in cytology-based cervical cancer screening ( 25 , 26 ). Our previous studies also showed that AI-based cytology was comparable and feasible to manual cytology in the general and referral populations ( 16 , 27 ).…”

Section: Discussionmentioning

confidence: 99%

“…This study further showed that the effectiveness of HPV strategies incorporating AI technology is comparable to that of incorporating manual cytology. Nonetheless, unlike in primary cytology screening ( 25 , 26 ), the role of AI in the reduction in manual work is limited in HPV-based strategies because many women who are HPV positive have abnormal cytology and need TBS classification by manual work. Further studies are needed to advance AI technology in the automated TBS classification of cytology.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

How Can a High-Performance Screening Strategy Be Determined for Cervical Cancer Prevention? Evidence From a Hierarchical Clustering Analysis of a Multicentric Clinical Study

Bao

Zhang

et al. 2022

Front. Oncol.

View full text Add to dashboard Cite

BackgroundThis study aimed to explore the cluster patterns of cervical cancer screening strategies for detecting high-grade precancerous lesions in terms of benefits, costs, and efficiency.MethodsA total of 2,065 referral women aged 25–64 years were recruited and underwent human papillomavirus (HPV) testing, liquid-based cytology with manual reading, and cytology with artificial intelligence (AI)-assisted reading. All women were assessed by colposcopy and histological examination. We formed 14 screening strategies based on primary cytology screening, primary HPV screening incorporating HPV-16/18 genotyping triage, cytology triage, or both, and co-testing. The primary outcomes were cervical intraepithelial neoplasia grade 2 or worse (CIN2+) and grade 3 or worse (CIN3+). The hierarchical clustering method was applied to multifaceted indicators, and then, the resulting clusters were described in terms of benefits, costs, efficiency, and their interaction. This study was registered (No. ChiCTR2000034131).ResultsThe relative sensitivity of HPV-based strategies compared with cytology alone with the threshold of LSIL+ ranged from 0.68 to 1.19 for CIN2+ detection and from 0.72 to 1.11 for CIN3+ detection, whereas the relative specificity ranged from 0.55 to 1.43 for CIN2+ detection and from 0.51 to 1.51 for CIN3+ detection. Five significant clusters according to the trade-off among benefits, costs, and efficiency were identified. The cluster including four primary HPV screening strategies showed the optimal balance. HPV testing with HPV-16/18 genotyping and AI-based cytology triage presented the optimal trade-off for CIN3+ detection relative to cytology alone in terms of relative sensitivity (1.06), relative specificity (0.72), colposcopies for 1 CIN3+ (3.7 vs. 3.1), a load of follow-up for women with HPV-positive and normal cytology (7.0% vs. 22.3%), and the work of manual cytology reading (35.1% vs. 100%).ConclusionsOur study provided clinical and methodological evidence on the choice of HPV-based screening strategies. The cluster including primary HPV screening with genotyping and cytology triage showed an optimal balance among benefit, cost, and efficiency.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

How Can a High-Performance Screening Strategy Be Determined for Cervical Cancer Prevention? Evidence From a Hierarchical Clustering Analysis of a Multicentric Clinical Study

Bao

Zhang

et al. 2022

Front. Oncol.

View full text Add to dashboard Cite

show abstract

“…developed an AI-assisted TBS (AIATBS) ( 21 ) diagnostic system, which showed higher sensitivity than the diagnosis by senior cytologists. The sensitivity of the AIATBS in detecting CIN was 94.74% in a clinical prospective validation ( 63 ).…”

Section: Applications Of Ai In Early Screening Of Cervical Cancermentioning

confidence: 99%

Artificial Intelligence in Cervical Cancer Screening and Diagnosis

Hou

Shen

Zhou³

et al. 2022

Front. Oncol.

View full text Add to dashboard Cite

Cervical cancer remains a leading cause of cancer death in women, seriously threatening their physical and mental health. It is an easily preventable cancer with early screening and diagnosis. Although technical advancements have significantly improved the early diagnosis of cervical cancer, accurate diagnosis remains difficult owing to various factors. In recent years, artificial intelligence (AI)-based medical diagnostic applications have been on the rise and have excellent applicability in the screening and diagnosis of cervical cancer. Their benefits include reduced time consumption, reduced need for professional and technical personnel, and no bias owing to subjective factors. We, thus, aimed to discuss how AI can be used in cervical cancer screening and diagnosis, particularly to improve the accuracy of early diagnosis. The application and challenges of using AI in the diagnosis and treatment of cervical cancer are also discussed.

show abstract

“…They reported a fabulous WSI classification performance but none of the cell segmentation/classification on external test set. Subsequently, Zhu et al [32] proposed an integrated cervical WSI recognition system, which includes 24 object detection CNNs and other 4 models. To complete the WSI classification task, they also created a new 24 classes to deal with the confusing cervical cancer subclasses, and arranged each object detection CNN for each class.…”

Section: B Detection From Whole Slide Imagementioning

confidence: 99%

Cervical Glandular Cell Detection from Whole Slide Image with Out-Of-Distribution Data

Wei¹,

Cheng²,

Cai³

et al. 2022

Preprint

View full text Add to dashboard Cite

Cervical glandular cell (GC) detection is a key step in computer-aided diagnosis for cervical adenocarcinomas screening. It is challenging to accurately recognize GCs in cervical smears in which squamous cells are the major. Widely existing Out-Of-Distribution (OOD) data in the entire smear leads decreasing reliability of machine learning system for GC detection. Although, the State-Of-The-Art (SOTA) deep learning model can outperform pathologists in preselected regions of interest, the mass False Positive (FP) prediction with high probability is still unsolved when facing such gigapixel whole slide image. This paper proposed a novel PolarNet based on the morphological prior knowledge of GC trying to solve the FP problem via a self-attention mechanism in eight-neighbor. It estimates the polar orientation of nucleus of GC. As a plugin module, PolarNet can guide the deep feature and predicted confidence of general object detection models. In experiments, we discovered that general models based on four different frameworks can reject FP in small image set and increase the mean of average precision (mAP) by 0.007 ∼ 0.015 in average, where the highest exceeds the recent cervical cell detection model 0.037. By plugging PolarNet, the deployed C++ program improved by 8.8% on accuracy of top-20 GC detection from external WSIs, while sacrificing 14.4 s of computational time. Code is available in https://github.com/Chrisa142857/PolarNet-GCdet.

show abstract

Hybrid AI-assistive diagnostic model permits rapid TBS classification of cervical liquid-based thin-layer cell smears

Cited by 47 publications

References 57 publications

How Can a High-Performance Screening Strategy Be Determined for Cervical Cancer Prevention? Evidence From a Hierarchical Clustering Analysis of a Multicentric Clinical Study

How Can a High-Performance Screening Strategy Be Determined for Cervical Cancer Prevention? Evidence From a Hierarchical Clustering Analysis of a Multicentric Clinical Study

Artificial Intelligence in Cervical Cancer Screening and Diagnosis

Cervical Glandular Cell Detection from Whole Slide Image with Out-Of-Distribution Data

Contact Info

Product

Resources

About