A processing strategy for automated Papanicolaou smear screening.

A multilayer processing strategy was developed for the automatic screening of conventionally prepared Papanicolaou smears. The processing stages include image segmentation, feature extraction, object classification and slide classification. Mathematical morphology functions were implemented in hardware with custom-built gate array processors for image segmentation. There were 68 features used for classifier training. In object classification we combined the evidential supports of a binary decision tree classifier and a multilayer perceptron classifier to achieve an integrated decision. In this feasibility study, 449 conventionally prepared cervical Papanicolaou smears were tested in a prototype research system between January and May 1991. The 95% confidence interval for the slide false-negative rate was 1-9%, and the 95% confidence interval for the slide sort rate was 45-55%. The estimated sort rate for clearly normal slides is within the range required for a cost-efficient screening system, and the estimated false-negative rate for premalignant and malignant smears is an improvement over published false-negative rates for human performance. Several performance improvement efforts are still under way. We expect that they will result in a vastly reduced slide false-negative rate.     

Cervical false negative cases detected by neural network-based technology. Critical review of cytologic errors

OBJECTIVE: To analyze the cytological errors made in the manual screening of cervical smears and to evaluate the usefulness of neural network-based technology (nnbt) in the detection of different kinds of errors. STUDY DESIGN: We reviewed 1,981 cervical smears by nnbt. Twelve false negatives (FNs) were detected and selected for study. The number of cell images showing atypical keratosis or atypical cells was evaluated on the monitor. The cellular features of the atypical cells (cellularity, cell type, nuclear and cytoplasmic changes, cellular arrangement and location on the slide) were analyzed by optic microscopy. Considering these qualitative and quantitative cytologic parameters and the diagnosis made by manual screening, we classified the errors into two groups: screening and interpretation related. RESULTS: Four FNs were screening errors. Five FNs were classified as errors of interpretation. In three cases the cause of the cytologic errors could not be ascertained. CONCLUSION: Our results confirm previous studies demonstrating that nnbt is useful for detecting screening errors. We also showed that it might be an adjunctive tool in the interpretation of abnormal cells, reducing the number of false negatives.     

A rapid screening classifier for diagnosing COVID-19

Rationale: Coronavirus disease 2019 (COVID-19) has caused a global pandemic. A classifier combining chest X-ray (CXR) with clinical features may serve as a rapid screening approach.Methods: The study included 512 patients with COVID-19 and 106 with influenza A/B pneumonia. A deep neural network (DNN) was applied, and deep features derived from CXR and clinical findings formed fused features for diagnosis prediction.Results: The clinical features of COVID-19 and influenza showed different patterns. Patients with COVID-19 experienced less fever, more diarrhea, and more salient hypercoagulability. Classifiers constructed using the clinical features or CXR had an area under the receiver operating curve (AUC) of 0.909 and 0.919, respectively. The diagnostic efficacy of the classifier combining the clinical features and CXR was dramatically improved and the AUC was 0.952 with 91.5% sensitivity and 81.2% specificity. Moreover, combined classifier was functional in both severe and non-serve COVID-19, with an AUC of 0.971 with 96.9% sensitivity in non-severe cases, which was on par with the computed tomography (CT)-based classifier, but had relatively inferior efficacy in severe cases compared to CT. In extension, we performed a reader study involving three experienced pulmonary physicians, artificial intelligence (AI) system demonstrated superiority in turn-around time and diagnostic accuracy compared with experienced pulmonary physicians.Conclusions: The classifier constructed using clinical and CXR features is efficient, economical, and radiation safe for distinguishing COVID-19 from influenza A/B pneumonia, serving as an ideal rapid screening tool during the COVID-19 pandemic.     

Expert consultation for cervical carcinoma smears. Reliability of selected-field videomicroscopy

OBJECTIVE: To evaluate the diagnostic accuracy of videomicroscopy image selection for expert consultation in cervical cytology. STUDY DESIGN: One hundred diagnostically difficult cervical cytologic smears were selected and rescreened by a general pathologist who chose, from each slide, four or five fields featuring abnormal cells. Video images were digitized and stored on a 512 x 512-pixel matrix using an image acquisition and transmission system. Five experts each reviewed 20 of the 100 cases, and a sixth reviewed all 100 cases. Diagnoses based on selected digitized images were compared to those based on conventional examination of whole slides. RESULTS: Intraobserver agreement was fair to excellent for all six experts (kappa value: 0.47-0.81); it was complete or acceptable in 68.4-85% of cases. Compared to the reference diagnosis, interobserver agreement was not significantly different whether cases were examined by screening the entire slide or by videomicroscopy of selected fields. The marked discordance in four cases concerned very small cells the significance of which was misinterpreted on videomicroscopy because of poor image quality due to lack of focus setting. CONCLUSION: This exploratory study showed that selection of videomicroscopy images seems as reliable as conventional examination of slides for expert consultation on diagnostically difficult cervical cytologic smear cases.     

Contextual analysis and intermediate cell markers enhance high-resolution cell image analysis for automated cervical smear diagnosis.

Until now, efforts to automate cervical smear diagnosis have focused on analyzing features of individual cells. In a complex specimen such as that obtained from a cervical scrape, diagnostically significant cells may not be adequately represented or may elude detection by the automated technology. An approach is needed that extracts additional quantitative information from cervical smears beyond what the cell-by-cell approach can provide. A new methodology, contextual analysis, was developed to extract global quantitative information about cells, cell clusters, and background debris. This pilot study was designed to compare the efficacy of contextual analysis with high-resolution, single cell analysis and the analysis of intermediate cell markers. Thirty-four samples prepared as monolayers and stained with the Feulgen-Thionin/Congo Red stain were measured. Contextual analysis alone was able to classify 91% of the smears correctly; single cell analysis classified 94% of the cells correctly; and the intermediate cell analysis correctly identified the smear diagnosis for 84% of the cells. When all three analysis methods were combined into a simple smear level classifier, the overall smear classification accuracy was improved over those obtained using the three methodologies alone.     

Potential of radial basis function neural networks in discriminating benign from malignant lesions of the lower urinary tract

OBJECTIVE: To investigate the potential value of morphometry and neural network tools for discriminating benign from malignant nuclei and lesions of the lower urinary tract. STUDY DESIGN: The study group consisted of 33 cases of lithiasis, 41 cases of inflammation, 66 cases of benign hyperplasia of the prostate, 4 cases of carcinoma in situ, 48 cases of grade 1 transitional cell carcinoma of the bladder (TCCB) and 123 cases of grade 2 and 3 TCCB. Images of routinely processed voided urine smears stained by the Giemsa technique were analyzed by a custom image analysis system. Analysis of the images gave a data set of features from 31,158 nuclei. A radial basis function (RBF)-type neural network was employed to discriminate benign from malignant nuclei, based on the extracted morphometric and textural features. Subsequently a second RBF classifier was employed to discriminate benign from malignant cases. The nuclei from 156 randomly selected cases (50% of total cases) was used as a training set, and the nuclei from the remaining 159 cases made up the test set. Similarly, in an attempt to discriminate at the patient level, the same 156 cases were used to train an RBF classifier; the remaining 159 cases were used for the test set. The cases used for training and testing the 2 classifiers (nuclear and patient level) were the same for the 2 kinds of classifiers. RESULTS: Application of the RBF classifier permitted the correct classification of 93.64% of benign nuclei and 85.61% of malignant, giving an overall accuracy of 84.45%. At the patient level the RBF classifier permitted an overall accuracy of 94.97%. These results were on the test sets. CONCLUSION: The role of nuclear morphologic features in the cytologic diagnosis of lower urinary tract alterations was confirmed by the results of this study. The observed overlap in feature space indicates that the nuclear characteristics do not form strictly separate clusters; that fact explains the difficulty morphologists have with reproducible identification of nuclei from the lower urinary tract. Application of RBF offers good classification at the nuclear and patient level and promises to become a powerful tool for everyday practice in the cytologic laboratory.     

Cytological screening for cervical cancer and human papillomavirus in general practice.

In a retrospective study of cervical screening in a general practice in Birmingham 156 out of 1913 smears taken over three years showed some abnormality. Smears from 65 women showed severe non-specific inflammation, and 91 women had various grades of cervical intraepithelial neoplasia, of whom 53 were aged under 30 and 13 over 40. Of 35 women with clinical evidence of human papillomavirus, 21 had normal results on cervical testing and 14 abnormal results. The incidence of genital warts among sexually active young people is growing, but the association of human papillomavirus with abnormal cervical smears is not clear. The efficacy of screening in the United Kingdom must be improved by actively encouraging younger patients to attend for regular screening.     

Feasibility of cytology-based cervical cancer screening in rural Cameroon

OBJECTIVE: To determine the prevalence of cervical cancer and its precursors in a rural population in Cameroon and to evaluate the feasibility of a cytology-based screening program in such areas. STUDY DESIGN: A prospective study was conducted in the rural town of Bafang. Following an advocacy campaign, 750 women were recruited. After a clinical examination, all women had a Pap smear with the Cervex Brush. Each sample had two preparations, conventional and liquid based. The conventional smears were interpreted in Bafang. Cytologically abnormal cases, those with clinical inflammation and/or macroscopic cervical lesions, had a colposcopic examination and directed biopsy. HSIL and colposcopically abnormal cases were treated with large loop excision of the transformation zone (LLETZ). The liquid-based preparations and histopathology were performed in Geneva and the results sent to Cameroon for patient follow-up. RESULTS: Mean age and parity of the women screened were 43.7 years and 7.8, respectively. The conventional smears showed 3.6% cervical abnormalities: 2% (15/740) ASCUS/LSIL and 1.6% (12/740) HSIL. The liquid-based preparations showed 12.6% (91/722) cervical abnormalities: 10.1% (73/722) ASCUS/LSIL and 2.5% (18/722) HSIL. Fifty percent of samples in both preparations showed evidence of inflammation. Histology was performed on 64 colposcopically directed punch biopsies and LLETZ specimens. The histologic diagnoses agreed with the cytologic findings in 60% (14/23) of conventional smears and 85% (12/14) of liquid-based preparations. CONCLUSION: There is a high rate of cervical intraepithelial neoplasia in the unscreened rural population of Cameroon. The situation is complicated by a high rate of cervical infection. A population-based cytologic screening program for cervical cancer would not be feasible in rural Cameroon because of high cost, low quality and limited technical facilities. Rural Africa requires an algorithm using a simple, low-cost technique of mass screening and an improved cytology service only to triage selected patients.     

Classification of complete blood count and haemoglobin typing data by a C4.5 decision tree,a naïve Bayes classifier and a multilayer perceptron for thalassaemia screening

This article presents the classification of blood characteristics by a C4.5 decision tree, a naïve Bayes classifier and a multilayer perceptron for thalassaemia screening. The aim is to classify eighteen classes of thalassaemia abnormality, which have a high prevalence in Thailand, and one control class by inspecting data characterised by a complete blood count (CBC) and haemoglobin typing. Two indices namely a haemoglobin concentration (HB) and a mean corpuscular volume (MCV) are the chosen CBC attributes. On the other hand, known types of haemoglobin from six ranges of retention time identified via high performance liquid chromatography (HPLC) are the chosen haemoglobin typing attributes. The stratified 10-fold cross-validation results indicate that the best classification performance with average accuracy of 93.23% (standard deviation = 1.67%) and 92.60% (standard deviation = 1.75%) is achieved when the naïve Bayes classifier and the multilayer perceptron are respectively applied to samples which have been pre-processed by attribute discretisation. The results also suggest that the HB attribute is redundant. Moreover, the achieved classification performance is significantly higher than that obtained using only haemoglobin typing attributes as classifier inputs. Subsequently, the naïve Bayes classifier and the multilayer perceptron are applied to an additional data set in a clinical trial which respectively results in accuracy of 99.39% and 99.71%. These results suggest that a combination of CBC and haemoglobin typing analysis with a naïve Bayes classifier or a multilayer perceptron is highly suitable for automatic thalassaemia screening.     

Rapid pre-screening of cervical smears as a method of internal quality control in a cervical screening programme

Objective:  To evaluate the performance of rapid pre-screening (RPS) as a method of internal quality control in the cytopathological examination of cervical smears for cervical cancer screening.
Methods:  The sample consisted of 6135 cervical smears submitted to RPS and routine screening (RS) methods. The smears classified as negative in RPS and RS were considered final diagnoses, and were not, therefore, submitted to any additional review. The smears identified as suspect or unsatisfactory according to RPS were analysed separately by two different cytologists irrespective of the diagnosis reached in RS. Smears considered abnormal or unsatisfactory at RS were also reviewed. When both cytologists issued concordant diagnoses, this was considered the final diagnosis. Discordant results were analysed by a third cytologist and a consensus meeting was held to define the final diagnosis.
Results:  Taking abnormalities detected by RS as the denominator, RPS had a sensitivity of 63.0% for the detection of all abnormal smears and 96.7% for high grade squamous intraepithelial lesion (HSIL). When compared with the final diagnosis, sensitivity of RPS for all abnormal smears was 74.9% and for HSIL 95.0%. Of the 529 abnormal smears confirmed in the final diagnosis, 2.15% were detected only by the RPS.
Conclusion:  RPS is an effective alternative method of internal quality control with high sensitivity for the detection of more severe lesions. It also permits monitoring of the laboratory rate of false-negative results, and allows constant evaluation of the performance both of the pre-screening and RS cytologists.     

Estimating the sensitivity of cervical cytology: errors of interpretation and test limitations

The objective of this study was to estimate: (i) the sensitivity of cytologists in recognizing abnormal smears; (ii) the sensitivity of cervical cytology as a method of detecting abnormal smears among those obtained in the presence of cervical intraepithelial neoplasia (CIN). Study subjects were 61 women with a histologically confirmed CIN identified through colpohistological and cytologic screening. For objective (i) new smears were taken from study subjects just before treatment, mixed with routine preparations, interpreted by unaware cytologists and then blindly reviewed by a group of three expert supervisors, who reached a consensus diagnosis. Cytologists classified as positive for squamous intraepithelial lesion (SIL) 30 of the 34 smears judged as positive by supervisors (100% of smears classified as high-grade and 67% of smears classified as low-grade SIL by the supervisors). Our approach, based on creating a set of smears with a high a priori probability of being positive, proved to be an efficient way of estimating errors of interpretation. For objective (ii), smears taken at the moment of diagnosis, just before biopsy, were also reviewed by the same supervisors. These CIN cases were identified among asymptomatic women independently of cytological findings and results are therefore not subject to verification bias. Among the 33 histological CINII/III, four (12%) smears had no atypical cells (three negatives and one unsatisfactory) at review. The same proportion was 26% (four negatives and one unsatisfactory) among the 19 histological CINI. No significant differences in smear content were found between the seven ‘false negatives’ and a sample of ‘true positives’ and ‘true negatives’ for a number of formal adequacy criteria (including presence of endocervical cells). Strong differences were found between positive smears taken just before biopsy and those taken just before treatment (in 11 women the first smear only was positive, while the opposite was never observed), suggesting an effect of punch biopsy in removing lesions.     

An assessment of partial rescreening as an internal quality control method for cervical smears

Partial screening was performed on 10 800 cervical smears, comprising 8640 filed negative and unsatisfactory smears and 2160 newly received smears prior to conventional screening. Each slide was screened for 30 s and those considered abnormal were reviewed by standard screening. Partial screening led to the detection of 27 additional infections and 44 additional cytological abnormalities. These detection rates are better than those obtained with the traditional method of rescreening only a proportion of smears. Amongst the smears partially screened before conventional screening, partial screening detected 37-66% of infections and 22-71% of cytological abnormalities. We recommend the use of partial rescreening of all negatively reported smears as a method of internal quality control in cervical cytology laboratories.     

Segmentation of cervical cell images.

A major problem in the automation of cervical cytology screening is the segmentation of cell images. This paper describes various standard segmentation methods plus one which determines a segmentation threshold based on the stability of the perimeter of the cell as the threshold is varied. As well as contour, certain structural information is used to decide upon the threshold which separates cytoplasm from the background. Once the cytoplasm threshold is found, cytoplasm and nucleus are separated by simple clustering into three groups, cytoplasm, folded cytoplasm and nucleus. These techniques have been tested on 1500 cervical cells that belong to one of eight normal classes and five abnormal classes. A minimum Mahalanobis distance classifier was used to compare results. Manually thresholded cells were classified correctly 66.0% of the time for the 13 class problem and 95.2% of the time on the two (normal-abnormal) class problem. The contour tracing technique was 52.9% and 90.0% correct, respectively.     

Quality control of cervical cytology in high-risk women. PAPNET system compared with manual rescreening

OBJECTIVE: To compare the effectiveness of the PAPNET System with conventional rescreening of negative cervical smears in a high-risk population. STUDY DESIGN: Three thousand ninety-seven negative cervical smears from women with past history of cervical abnormalities were rescreened manually and with the PAPNET System. There were two reviews of PAPNET images: the first by two cytotechnologists with limited exposure to the instrument, and the second, limited to smears with discrepant diagnoses, by an expert in the use of the system. The remaining discrepant smears were submitted to a blinded microscopic review by a third party. The a priori consensus diagnosis was arbitrarily established when the result of two of the three reviews--manual, PAPNET and the independent third review--were concordant. The results of rescreening were compared with available biopsies. RESULTS: On manual rescreening of the 3,097 smears, 2,901 (93.66%) were reported as negative and 170 (5.49%) as abnormal. On the first PAPNET review, 2,938 (94.87%) were reported as negative and 150 (4.84%) as abnormal. There were 144 smears with discrepant diagnoses. After the second PAPNET review of these discrepant smears, the agreement between manual and PAPNET rescreening rose from 94.27% to 95.58%. A final, blinded review of 89 residual discrepant smears was used to establish consensus diagnoses. The diagnoses made by PAPNET-assisted rescreening agreed much better with the consensus diagnoses than did manual rescreening (Kappa = .61 vs. Kappa = -.32, P < .001). When compared with the results of 50 available biopsies, PAPNET-assisted rescreening also had a somewhat lower false negative rate (sensitivity 58.82% vs. 41.18%, P = .17) and a statistically significant lower false positive rate (specificity 63.64% vs. 36.36%, P = .01). CONCLUSION: PAPNET-assisted rescreening, when carried out by an experienced person, is more efficient than manual rescreening.     

Variation in the assessment of adequacy in cervical smears

OBJECTIVE: To assess the interobserver reproducibility of the diagnosis of 'adequacy' of cervical smears according to the Bethesda System criteria in cervical smears. STUDY DESIGN: 358 cervical smears were obtained from three Italian cytopathological centres in 1998-99. All centres provided consecutively collected smears. The cervical smears were independently and blindly assessed by four cytologists.The screening was performed using a 10x objective and an additional evaluation of the percentage of cellularity was performed using a 4x objective. RESULTS: The proportion of smears assessed by the four cytologists as 'adequate' ranged from 60% to 70%, the proportion of 'satisfactory for evaluation but limited by' ranged from 27% to 38%, and the proportion of 'inadequate smears' ranged from 2% to 4%. Full agreement in the assessment of smear adequacy was observed in 311 slides and disagreement was observed only in 47. The category 'inadequate smear' was less reliable than the other two; however, the kappa value observed was acceptable. CONCLUSION: The present study shows that it is possible to achieve a high reproducibility in the assessment of smear adequacy, at least among expert cytologists who follow the Bethesda System criteria strictly.     

Cervicovaginal cytology in an indigent population. Comparison of results for 1964, 1981 and 1989

To evaluate the usefulness of cervicovaginal cytology in decreasing the incidence of cervical carcinoma in an indigent population, the cytologic findings from 10,000 consecutive smears in 1964 (when cytology screening started) were compared to the results of 10,000 consecutive smears in 1981 and 1989. There was a marked (statistically significant) decrease in invasive cervical squamous carcinoma at all ages between the first and later periods. Squamous carcinoma in situ showed a significant decrease beginning in patients under 40 in 1981. The number of atypias and mild dysplasias showed a proportional increase, from 2% in 1964 to 13.4% in 1981 to 21.8% in 1989, predominantly in young patients. These results reaffirm that cervicovaginal cytology remains the most inexpensive and effective diagnostic tool for the elimination of cervical cancer.     

Cervical cytology screening: experience of a general hospital.

At Henderson General Hospital, Hamilton, a program was introduced whereby cervical smears were taken routinely for cytologic study from all women admitted aged 17 years or older. The procedure was performed by a specially trained nurse. In a 5-year period 53% of eligible patients were screened. Of these, 32% had not had a cervical smear taken before. In 7681 smears nine instances of invasive disease were discovered: three of the cervix, three of the endometrium and three metastatic. There were 20 cases of carcinoma in situ and 2 of severe dysplasia. Evidence of infection was present in a high percentage of the smears. Hospital admission affords an excellent opportunity of applying this valuable screening procedure.     

Significance of high-risk human papillomavirus detection by polymerase chain reaction in primary cervical cancer screening

The purposes of this study were to evaluate the incidence of high-risk human papillomavirus (HPV) infection by polymerase chain reaction (PCR) and to assess its diagnostic usefulness in primary cervical screening. PCR testing for HPV type 16, 18, 31 and 33 was performed on 1305 specimens obtained during routine cervical cancer screening. We analysed the concurrent cervical smears and biopsy, and correlated them with the HPV infection status. We also evaluated histologically-proven cases with ASCUS smears according to HPV infection. HPV DNA was identified in eight (0.7%) of 1144 cytologically normal patients; nine (10.5%) of 86 ASCUS; seven (25.0%) of 28 LSIL; 26 (78.8%) of 33 HSIL; and in all of three squamous cell carcinomas (SCC). HPV positivity was significantly associated with cytohistological diagnosis for HSIL of more. In addition, HPV-positive ASCUS cases were found to be associated with histological abnormality rather than HPV-negative. The results indicate that high-risk HPV testing by PCR could be a useful adjunct tool for Pap smear in primary cervical screening. The combination of Pap smear and high-risk HPV testing by PCR might reduce unnecessary colposcopy-guided biopsy of women with cytological diagnosis of ASCUS.     

Rapid rescreening of cervical smears as a quality control method in a high-risk population

OBJECTIVE: Cancer of the cervix is one of the commonest cancers in South Africa. Accurate cytological diagnosis is one of the prerequisites for an effective cervical screening programme and requires the implementation of appropriate quality assurance modalities. This study was undertaken to determine if rapid review of reportedly negative cervical smears is a useful internal quality assurance modality in an unscreened population with very high rates of cervical carcinoma. METHOD: Approximately 26% of all cervical smears received at the study institution between 1 January 1998 and 31 December 2003, and initially reported as negative or inadequate, underwent rapid review. RESULTS: A total of 62,866 (26%) cervical smears out of 241,796 reportedly negative or inadequate cervical smears underwent rapid review. An amended report was sent out in 373 (0.59%) of these 62,866 cervical smears. This included 101 cases of high-grade squamous intraepithelial lesion (HSIL) and high-grade atypical squamous cells (ASC-H), 143 low-grade squamous intraepithelial lesions, 54 atypical squamous cells of undetermined significance (ASC-US) and 33 atypical glandular cells that were not reported initially. The false-negative proportion for HSIL and ASC-H (combined) in this study was 5.76%. No squamous cell carcinomas were diagnosed on rapid review but one patient with HSIL/ASC-H on review had squamous cell carcinoma on biopsy. Three cytotechnologists had a lower sensitivity of primary screening and required retraining. CONCLUSIONS: Rapid review is beneficial as an internal quality assurance modality in an unscreened high-risk population and increases the detection of women with significant cervical lesions requiring treatment. The relatively low cost of rapid review compared with other rescreening modalities makes this an attractive option in low resource settings.     

Computerized morphometric discrimination between normal and tumoral cells in oral smears

The oral exfoliative cytology allows a quick and fairly accurate assessment of suspicious lesions of the oral cavity. Within this context, our paper proposes a quantitative approach, focusing on the construction of a classifier for detecting the presence of the tumoral cells on oral smears. The design of the classifier relies on a detailed computerized analysis of the individual morphometric features exhibited by two large known populations of normal and tumoral cells, respectively; the digital image processing was performed in the Zeiss KS400 environment. The classifier was implemented as a neural network with step activation function, whose parameters were obtained from an adequate training, based on the nuclear and cytoplasmic areas of the cells belonging to the two populations. Our procedure based on this classifier was meant to operate by identifying the tumoral or normal nature of any cell randomly selected from a smear. To identify the nature of an arbitrary cell, its nuclear and cytoplasmic areas are presented at the input of the classifier. The classification procedure was tested on several smears, and the results coincided with the pathological diagnosis in all the considered cases. The performances of our approach are discussed in comparison with other analytical methods previously reported in oral exfoliative cytology. These discussions emphasize the role of numerical information exploited for the classifier design, concluding that the individual morphometric features are more meaningful than the global characterization of smears by mean values.