Long-term observational approach in women with histological diagnosis of cervical low-grade squamous intraepithelial lesion: an Italian multicentric retrospective cohort study

Application of EfficientNet-B0 and GRU-based deep learning on classifying the colposcopy diagnosis of precancerous cervical lesions

BACKGROUND

Colposcopy is indispensable for the diagnosis of cervical lesions. However, its diagnosis accuracy for high-grade squamous intraepithelial lesion (HSIL) is at about 50%, and the accuracy is largely dependent on the skill and experience of colposcopists. The advancement in computational power made it possible for the application of artificial intelligence (AI) to clinical problems. Here, we explored the feasibility and accuracy of the application of AI on precancerous and cancerous cervical colposcopic image recognition and classification.

METHODS

The images were collected from 6002 colposcopy examinations of normal control, low-grade squamous intraepithelial lesion (LSIL), and HSIL. For each patient, the original, Schiller test, and acetic-acid images were all collected. We built a new neural network classification model based on the hybrid algorithm. EfficientNet-b0 was used as the backbone network for the image feature extraction, and GRU(Gate Recurrent Unit)was applied for feature fusion of the three modes examinations (original, acetic acid, and Schiller test).

RESULTS

The connected network classifier achieved an accuracy of 90.61% in distinguishing HSIL from normal and LSIL. Furthermore, the model was applied to "Trichotomy", which reached an accuracy of 91.18% in distinguishing the HSIL, LSIL and normal control at the same time.

CONCLUSION

Our results revealed that as shown by the high accuracy of AI in the classification of colposcopic images, AI exhibited great potential to be an effective tool for the accurate diagnosis of cervical disease and for early therapeutic intervention in cervical precancer.

Association of Human forkhead box protein 3 (FOXP3) gene polymorphisms with idiopathic recurrent pregnancy loss among Kazakhstani women

BACKGROUND

Recurrent pregnancy loss (RPL) is major pregnancy complication, with poorly defined cause.Forkhead Box P3 (FOXP3) is a transcription factor that supports Treg activation and development and attenuates immune responses. As FOXP3 production is genetically determined, we tested the association of FOXP3 gene variants with RPL.

METHODS

A retrospective case-control study, performed between April 2019 and February 2020. Study subjects comprised 62 RPL cases and 60 control women. Genotyping of the four FOXP3 variants rs2294021 (T > C), rs2232365 (G > A), rs3761548 (C > A), and rs141704699 (C > T) was done by real-time PCR, with defined clusters. Logistic odds ratios (ORs) of RPL risk were estimated with 95% confidence interval (CI) after adjustment; statistical significance set at P < 0.05.

RESULTS

Minor allele frequency (MAF) of rs2294021 was significantly lower [P < 0.001; OR(95% CI) = 0.25(0.11-0.55)], while rs2232365 MAF was significantly higher [P = 0.045; OR(95% CI) = 1.85(1.05-3.28)] in cases, hence assigning RPL-protection and -susceptibility to these variants, respectively. Increased RPL risk was seen in rs2232365 homozygous minor allele carrying genotype [OR(95% CI) = 5.14(1.01-26.15)], while reduced RPL risk was noted in rs2294021 heterozygous [OR(95% CI) = 0.30(0.11-0.80)], and homozygous minor allele [OR(95% CI) = 0.10(0.01-0.83)] genotype carriers. Moderate linkage disequilibrium analysis was seen between the tested variants. Increased frequency of TACC, and reduced frequency of CGAC haplotypes were seen in RPL cases when compared to controls, thereby assigning RPL susceptibility and protection to these haplotypes, respectively.

CONCLUSION

These results suggest that FOXP3 variants and haplotypes are associated with idiopathic RPL, suggesting the likely contribution of Treg to RPL.

The Natural History of Cervical Intraepithelial Neoplasia Grades 1, 2, and 3: A Systematic Review and Meta-analysis

OBJECTIVE

The aim of the study was to obtain an updated overview of regression, persistence, and progression rates of conservatively managed cervical intraepithelial neoplasia grade 1 (CIN 1)/CIN 2/CIN 3.

METHODS

Data sources were MEDLINE, Embase, and Cochrane (January 1, 1973-April 14, 2020). Two reviewers extracted data and assessed risk of bias. To estimate outcome rates, we pooled proportions of the individual study results using random-effects meta-analysis, resulting in point estimates and corresponding 95% CIs. Heterogeneity was quantified by the I2 and τ2 measures.

RESULTS

Eighty-nine studies were included, 63 studies on CIN 1 (n = 6,080-8,767), 42 on CIN 2 (n = 2,909-3,830), and 7 on CIN 3 (n = 245-351). The overall regression, persistence, and progression to CIN 2 or worse and CIN 3 or worse rates for women with conservatively managed CIN 1 were 60% (95% CI = 55-65, I2 = 92%), 25% (95% CI = 20-30, I2 = 94%), 11% (95% CI = 8-13, I2 = 89%), and 2% (95% CI = 1-3, I2 = 82%), respectively. The overall regression, persistence, and progression rates for CIN 2 were 55% (95% CI = 50-60, I2 = 85%), 23% (95% CI = 19-28, I2 = 83%), and 19% (95% CI = 15-23, I2 = 88%), respectively. Finally, for CIN 3, these were 28% (95% CI = 17-41, I2 = 68%), 67% (95% CI = 36-91, I2 = 84%), and 2% (95% CI = 0-25, I2 = 95%), respectively. Cervical intraepithelial neoplasia grade 2 regression was significantly higher in women 30 years or younger and high-risk human papillomavirus-negative women (66%, 95% CI = 62-70, I2 = 76%; 94%, 95% CI = 84-99, I2 = 60%). Only 2/7,180 (0.03%) and 10/3,037 (0.3%) of the CIN 1 and CIN 2 cases progressed to cervical cancer.

CONCLUSIONS

Most CIN 1/CIN 2 will regress spontaneously in less than 24 months, with the highest rates in high-risk human papillomavirus-negative and young women, whereas progression to cancer is less than 0.5%. Conservative management should be considered, especially in fertile women and with expected high compliance. Given the heterogeneity in regression rates of high-grade histology, this should be classified as CIN 2 or CIN 3 to guide management.

Cost-effectiveness of HPV-based cervical screening based on first year results in the Netherlands: a modelling study

Objective

We aim to compare the cost‐effectiveness of the old cytology programme with the new high‐risk human papillomavirus (hrHPV) screening programme, using performance indicators from the new Dutch hrHPV screening programme.

Design

Model‐based cost‐effectiveness analysis.

Setting

The Netherlands.

Population

Dutch 30‐year‐old unvaccinated females followed up lifelong.

Methods

We updated the microsimulation screening analysis (MISCAN) model using the most recent epidemiological and screening data from the Netherlands. We simulated both screening programmes, using the screening behaviour and costs observed in each programme. Sensitivity analyses were performed on screening behaviour, utility losses and discount rates.

Main outcome measures

Cervical cancer incidence and mortality rates, number of screening tests and repeat tests, colposcopy referrals by lesion grade, costs from a societal perspective, quality‐adjusted life years (QALYs) gained and cost‐effectiveness.

Results

The new Dutch cervical cancer screening programme decreased the cervical cancer mortality by 4% and the incidence by 1% compared with the old programme. Colposcopy referrals of women without cervical intra‐epithelial neoplasia grade 2 or worse, increased by 172%, but 13% more QALYs were still achieved. Total costs were reduced by 21%, mainly due to fewer screening tests. Per QALY gained, the hrHPV programme cost 46% less (€12,225) than the cytology programme (€22,678), and hrHPV‐based screening remained more cost‐effective in all sensitivity analyses.

Conclusions

The hrHPV‐based screening programme was found to be more effective and cost‐effective than the cytology programme. Alternatives for the current triage strategy should be considered to lower the number of unnecessary referrals.

Tweetable abstract

First results after implementation confirm that HPV screening is more cost‐effective than cytology screening.

First results after implementation confirm that HPV screening is more cost‐effective than cytology screening.

Switches of SOX17 and SOX2 expression in the development of squamous metaplasia and squamous intraepithelial lesions of the uterine cervix

AIMS

The dynamics and topographical distribution of SOX17 and SOX2 expression was studied in the transformation zone (TZ) of the uterine cervix. This TZ is a dynamic area where switches from glandular into squamous epithelium can be recognized, new squamocolumnar junctions are formed, and premalignant lesions originate. SOX17 and SOX2 show mutually exclusive expression patterns in the normal uterine cervix, with SOX2 being exclusively found in squamous epithelium, while SOX17 is detected in endocervical columnar cells and reserve cells.

METHODS AND RESULTS

Normal cervices and squamous intraepithelial lesions (SIL) were studied with immunohistochemistry, methylation of SOX17, human papilloma virus (HPV) genotyping, and in situ hybridization. In the TZ squamous metaplasia originating from these reserve cells can still show SOX17 expression, while also remnants of SOX17-positive immature metaplasia can be recognized in the normal squamous epithelium. SOX17 expression is gradually lost during maturation, resulting in the exclusive expression of SOX2 in the majority of (SIL). This loss of SOX17 expression is independent of methylation of the CpG island in its promotor region. HPV can be detected in SOX17-positive immature metaplastic regions in the immediate vicinity of SOX2-positive SIL, suggesting that switches in SOX17 and 2 expression can occur upon HPV infection.

CONCLUSIONS

This switch in expression, and the strong association between the distribution of reserve cells and squamous areas within the columnar epithelium in the TZ, suggests that reserve cell proliferations, next to basal cells in the squamous epithelium, are potential targets for the formation of squamous lesions upon viral infection.