Less medical intervention after sharp demarcation of Grade 1-2 cervical intraepithelial neoplasia smears by neural network screening.

BACKGROUND: Neural network technology has been used for the daily screening of cervical smears in The Netherlands since 1992. The authors believe this method might have the potential to demarcate diagnoses of Grade 1-2 cervical intraepithelial neoplasia (CIN 1-2). METHODS: Of 133,196 women who were screened between 1992-1995, there were 2236 CIN 1-2 smears; 1128 of which were detected by means of neural network screening (NNS) (n = 83,404 women) and 1108 of which were diagnosed by conventional screening (n = 49,792 women). Cytologic and clinical outcomes (first cytologic or histologic follow-up diagnosis) were retrieved for all the women in the study population (n = 1920). Stratification based on clinical outcome resulted in the cases being grouped as overdiagnosed, concordant, or underdiagnosed. The smears were performed by general practitioners, whereas the biopsies were obtained by gynecologists. RESULTS: The prevalence rate for CIN 1-2 was 1.15% (95% confidence interval [95% CI], 1.08-1.23%) for NNS and 1.92% (95% CI, 1.80-2.04%) for conventional diagnosis (P < 0.001). Concordance with histology was significantly higher for NNS (53.9%; 95% CI, 50.7-57.0%) compared with conventional screening (29.2%; 95% CI, 26.4-32.2%). In addition, overdiagnosis was significantly lower for cases diagnosed by NNS (39.4%; 95% CI, 36.3-42.4%) compared with cases diagnosed by conventional screening (62.4%; 95% CI, 59.3-65.5%). CONCLUSIONS: Neural network-based screening can lead to fewer women being burdened unnecessarily with a cytologic diagnosis of CIN 1-2 by resulting in a sharp demarcation in these diagnoses and a corresponding reduction in unnecessary medical interventions. [See editorial on pages 171-172, this issue.]

Neural network-based screening (NNS) in cervical cytology: no need for the light microscope?

Neural network-based screening (NNS) of cervical smears can be performed as a so-called "hybrid screening method," in which parts of the cases are additionally studied by light microscope, and it can also be used as "pure" NNS, in which the cytological diagnosis is based only on the digital images, generated by the NNS system. A random enriched sample of 985 cases, in a previous study diagnosed by hybrid NNS, was drawn to be screened by pure NNS. This study population comprised 192 women with (pre)neoplasia of the cervix, and 793 negative cases. With pure NNS, more cases were recognized as severely abnormal; with hybrid NNS, more cases were cytologically diagnosed as low-grade. For a threshold value > or = HSIL (high-grade squamous intraepithelial lesions), the areas under the receiver operating characteristic (ROC) curves (AUC) were 81% (95% CI, 75-88%) for pure NNS vs. 78% (95% CI, 75-81%) for hybrid NNS. For low-grade squamous intraepithelial lesions (LSIL), the AUC was significantly higher for hybrid NNS (81%; 95% CI, 77-85%) than for pure NNS (75%; 95% CI, 70-80%). Pure NNS provides optimized prediction of HSIL cases or negative outcome. For the detection of LSIL, light microscopy has additional value.

Cytological recognition of invasive squamous cancer of the uterine cervix: comparison of conventional light-microscopical screening and neural network-based screening.

Cytologic recognition of invasive or microinvasive cancer of the uterine cervix may present substantial difficulties. In this study, we compared conventional light-microscopical screening of 109,104 cervical smears and neural network-based screening (NNS) of 245,527 smears, all obtained by the spatula-Cytobrush method. Two populations of Dutch women were included in the study: those receiving smears within the framework of the Dutch population screening program ("routine smears") and those receiving smears for other reasons, discussed in the text ("interval smears"). There were 71 smears, from an equal number of biopsy-confirmed invasive squamous carcinomas, 28 of which were microinvasive. The "interval smears" yielded a statistical valid higher prevalence of invasive cancer than "routine smears." Except for 5 smears that contained no evidence of abnormality ("sampling errors"), no false-negative errors occurred in the 52 NNS cases, whereas 4 such errors occurred in the 19 conventionally screened cases. By measuring the amount of cancerous material present in each smear (mapping), it could be documented that NNS was effective even in smears with a small number of cancer cells, whereas the 4 conventional false-negative screening errors occurred in smears of this type. The study showed that cells derived from invasive cancer of the cervix may have large bland nuclei that do not fit the images commonly associated with squamous cancer cells. Neural network-based screening of cervical smears was more effective than conventional screening in the diagnosis of invasive squamous cancer of the uterine cervix.

Potentially difficult smears of women with squamous cell carcinoma pose fewer problems when PAPNET is used for primary screening.

The diagnosis of squamous cell carcinoma (SCC) on a cervical smear is often far from easy. This study reports the analysis of 40 true-positive SCC smears detected in primary PAPNET screening and eight false-negative (FN) conventionally screened smears. All FN cases contained sparse abnormal material (< 10% of the slide). In these potentially difficult cases the diagnosis on the PAPNET images was not hard. Statistical analysis of the quantitative data indicated that the PAPNET images of the FN cases and the true-positive cases differed in some aspects. PAPNET highlighted the importance of background information (old blood, fibrin and necrosis). In addition, all FN smears contained cancer cells in the PAPNET images, allowing a correct diagnosis.

New paradigm for ASCUS diagnosis using neural networks.

It was tested whether it was possible to reduce the atypical squamous cells of undetermined significance (ASCUS) scores in a meaningful way by exploiting the cells selected by the neural networks of the PAPNET system. For this test, 2,000 routine smears were screened once by means of PAPNET and once conventionally in a laboratory in Amsterdam. From these 2,000 smears, 168 were diagnosed as ASCUS. In the second phase of the study, the diagnosis was based solely on the PAPNET images, and in addition, cases with immature cells (bare nuclei and cells with very little cytoplasm) in the PAPNET images were classified as ASCUS. Although, in this second phase, 75.6% of the cases were revised to negative, the cases with positive follow-up were all still classified as ASCUS. The negative predictive value remained at 100%, whereas the positive predictive value increased from 14.3 to 30%. By using the new paradigm (focusing on immature cells selected by the neural networks) for routine primary PAPNET screening in a laboratory in Leiden, the ASCUS scores were reduced from 10% (June of 1996) to 1.0% (early 1998), with promising follow-up results for the first half of 1997.