Enhancing colposcopy training using a widely accessible digital education tool in China

AI-Based Identification Method for Cervical Transformation Zone Within Digital Colposcopy: Development and Multicenter Validation Study

BACKGROUND

In low- and middle-income countries, cervical cancer remains a leading cause of death and morbidity for women. Early detection and treatment of precancerous lesions are critical in cervical cancer prevention, and colposcopy is a primary diagnostic tool for identifying cervical lesions and guiding biopsies. The transformation zone (TZ) is where a stratified squamous epithelium develops from the metaplasia of simple columnar epithelium and is the most common site of precancerous lesions. However, inexperienced colposcopists may find it challenging to accurately identify the type and location of the TZ during a colposcopy examination.

OBJECTIVE

This study aims to present an artificial intelligence (AI) method for identifying the TZ to enhance colposcopy examination and evaluate its potential clinical application.

METHODS

The study retrospectively collected data from 3616 women who underwent colposcopy at 6 tertiary hospitals in China between 2019 and 2021. A dataset from 4 hospitals was collected for model conduction. An independent dataset was collected from the other 2 geographic hospitals to validate model performance. There is no overlap between the training and validation datasets. Anonymized digital records, including each colposcopy image, baseline clinical characteristics, colposcopic findings, and pathological outcomes, were collected. The classification model was proposed as a lightweight neural network with multiscale feature enhancement capabilities and designed to classify the 3 types of TZ. The pretrained FastSAM model was first implemented to identify the location of the new squamocolumnar junction for segmenting the TZ. Overall accuracy, average precision, and recall were evaluated for the classification and segmentation models. The classification performance on the external validation was assessed by sensitivity and specificity.

RESULTS

The optimal TZ classification model performed with 83.97% classification accuracy on the test set, which achieved average precision of 91.84%, 89.06%, and 95.62% for types 1, 2, and 3, respectively. The recall and mean average precision of the TZ segmentation model were 0.78 and 0.75, respectively. The proposed model demonstrated outstanding performance in predicting 3 types of the TZ, achieving the sensitivity with 95% CIs for TZ1, TZ2, and TZ3 of 0.78 (0.74-0.81), 0.81 (0.78-0.82), and 0.8 (0.74-0.87), respectively, with specificity with 95% CIs of 0.94 (0.92-0.96), 0.83 (0.81-0.86), and 0.91 (0.89-0.92), based on a comprehensive external dataset of 1335 cases from 2 of the 6 hospitals.

CONCLUSIONS

Our proposed AI-based identification system classified the type of cervical TZs and delineated their location on multicenter, colposcopic, high-resolution images. The findings of this study have shown its potential to predict TZ types and specific regions accurately. It was developed as a valuable assistant to encourage precise colposcopic examination in clinical practice.

Assessing colposcopy competencies in medically underserved communities: a multi-center study in China

BACKGROUND

Colposcopy plays an essential role in diagnosing cervical lesions and directing biopsy; however, there are few studies of the capabilities of colposcopists in medically underserved communities in China. This study aims to fill this gap by assessing colposcopists' competencies in medically underserved communities of China.

METHODS

Colposcopists in medically underserved communities across China were considered eligible to participate. Assessments involved presenting participants with 20 cases, each consisting of several images and various indications. Participants were asked to determine transformation zone (TZ) type, colposcopic diagnoses and to decide whether biopsy was necessary. Participants are categorized according to the number of colposcopic examinations, i.e., above or below 50 per annum.

RESULTS

There were 214 participants in this study. TZ determination accuracy was 0.47 (95% CI 0.45,0.49). Accuracy for colposcopic diagnosis was 0.53 (95% CI 0.51,0.55). Decision to perform biopsies was 0.73 accurate (95% CI 0.71,0.74). Participants had 0.61 (95% CI 0.59,0.64) sensitivity and a 0.80 (95% CI 0.79,0.82) specificity for detecting high-grade lesions. Colposcopists who performed more than 50 cases were more accurate than those performed fewer across all indicators, with a higher sensitivity (0.66 vs. 0.57, p = 0.001) for detecting high-grade lesions.

CONCLUSIONS

In medically underserved communities of China, colposcopists appear to perform poorly at TZ identification, colposcopic diagnosis, and when deciding to biopsy. Colposcopists who undertake more than 50 colposcopies each year performed better than those who perform fewer. Therefore, colposcopic practice does improve through case exposure although there is an urgent need for further pre-professional and clinical training.

Genotype, cervical intraepithelial neoplasia, and type-specific cervical intraepithelial neoplasia distributions in hrHPV+ cases referred to colposcopy: A multicenter study of Chinese mainland women

To investigate age and type-specific prevalences of high-risk human papillomavirus (hrHPV) and cervical intraepithelial neoplasia (CIN) in hrHPV+ women referred to colposcopy. This is a retrospective, multicenter study. Participants were women referred to one of seven colposcopy clinics in China after testing positive for hrHPV. Patient characteristics, hrHPV genotyping, colposcopic impressions, and histological diagnoses were abstracted from electronic records. Main outcomes were age-related type-specific prevalences associated with hrHPV and CIN, and colposcopic accuracy. Among 4419 hrHPV+ women referred to colposcopy, HPV 16, 52, and 58 were the most common genotypes. HPV 16 prevalence was 39.96%, decreasing from 42.57% in the youngest group to 30.81% in the eldest group. CIN3+ prevalence was 15.00% and increased with age. As lesion severity increases, HPV16 prevalence increased while the prevalence of HPV 52 and 58 decreased. No age-based trend was identified with HPV16 prevalence among CIN2+, and HPV16-related CIN2+ was less common in women aged 60 and above (44.26%) compared to those younger than 60 years (59.61%). Colposcopy was 0.73 sensitive at detecting CIN2+ (95% confidence interval[CI]: 0.71, 0.75), with higher sensitivity (0.77) observed in HPV16+ women (95% CI: 0.74, 0.80) compared to HPV16- women (0.68, 95% CI: 0.64, 0.71). Distributions of hrHPV genotypes, CIN, and type-specific CIN in Chinese mainland hrHPV+ women referred to colposcopy were investigated for the first time. Distributions were found to be age-dependent and colposcopic performance appears related to HPV genotypes. These findings could be used to improve the management of women referred to colposcopy.

Chinese colposcopists' attitudes toward the colposcopic artificial intelligence auxiliary diagnostic system (CAIADS): A nation-wide, multi-center survey

Objective

The objective of this study was to assess the attitudes toward the Colposcopic Artificial Intelligence Auxiliary Diagnostic System (CAIADS) of colposcopists working in mainland China.

Methods

A questionnaire was developed to collect participants' sociodemographic information and assess their awareness, attitudes, and acceptance toward the CAIADS.

Results

There were 284 respondents from 24 provinces across mainland China, with 55% working in primary care institutions. Participant data were divided into two subgroups based on their colposcopy case load per year (i.e. ≥50 cases; <50 cases). The analysis showed that participants with higher loads had more experience working with CAIADS and were more knowledgeable about CAIADS and AI systems. Overall, in both groups, about half of the participants understood the potential applications of big data and AI-assisted diagnostic systems in medicine. Although less than one-third of the participants were knowledgeable about CAIADS and its latest developments, more than 90% of the participants were open with the idea of using CAIADS.

Conclusions

While a related lack of acknowledgement of CAIADS exists, the participants in general had an open attitude toward CAIADS. Practical experience with colposcopy or CAIADS contributed to participants' awareness and positive attitudes. The promotion of AI tools like CAIADS could help address regional health inequities to improve women's well-being, especially in low- and middle-income countries.