Biomedical subjects

L Kuan

Publications and source records attributed to L Kuan.

5 recordsLinked to original sources

PubMed1998

The conditional binomial test revisited for clinical trials.

McNemar's test is used to test the hypothesis that one treatment is better than another in a matched-pair design for binary outcomes. The conditional binomial test in such a matched-pair design is the exact McNemar test. However, in many clinical trials, one wants to establish equivalency between two treatments. We discuss how to use a conditional binomial test to establish equivalency between two treatments in a matched-pair design. Sample size and power determination for each conditional binomial test are calculated. Some statistical properties of the tests are analyzed through Monte Carlo simulation.

Binomial Distribution↗

PubMed1997

Effectiveness of automated cervical cytology rescreening using the AutoPap 300 QC System.

This study assesses the performance of the AutoPap 300 QC System in identifying false-negative (FN) smears in a slide population previously screened as normal and compares the detection rate to that achieved with a random rescreen of the same slide population. A total of 1,840 "normal" smears were rescreened both manually and by the AutoPap 300 QC System. Overall, a total of 7 FN slides were detected. At QC selection rates of 30% and 20% the device achieved sensitivities for detection of FN smears of 57.19% (4/7) and 42.8% (3/7), respectively. This represents a three- to fourfold enrichment in the number of FN smears over that obtained by a random rescreen of a similar proportion of cases. None of the FN slides were identified by either method at a 10% rescreening rate. The ability of the device to detect slides previously classified as abnormal (n = 139) and FN (n = 40) was also studied. The overall sensitivity to abnormal smears at QC selection rates of 10%, 20%, and 50% was 61.9%, 77.0%, and 94.2%, respectively. Improved sensitivity to smears classified as LSIL or worse (n = 112) was obtained for corresponding selection rates (61.6%, 75.9%, and 93.8%). Sensitivity to FN slides classified as LSIL or worse (n = 17) for QC selection rates of 10%, 20%, and 50% was 29.4%, 70.6%, and 88.2%, respectively. The sensitivity and specificity of the device to an adequate squamous and endocervical cell component was also determined. At predetermined thresholds, the overall sensitivity to slides with an inadequate squamous cell component (n = 55) and to those smears with an endocervical cell component (n = 1.587) was 81.8%, and 82.7% respectively. The study demonstrated that the AutoPap 300 QC System is superior to human random rescreen for the identification of FN smears although only a marginal improvement was noted due to the small sample. Further studies are required using a larger number of smears to fully assess the value of the device in quality control mode. The device also has the potential to improve the accuracy of specimen adequacy determinations and to serve as a useful adjunct to existing quality control measures designed to monitor individual performance and reporting accuracy.

Automation↗

PubMed1993

Grading cervical dysplasia with AgNORs using a semiautomated image analysis system.

Colposcopic biopsies were classified according to previously established criteria by a group of three pathologists interested in cervical pathology. Ten cases were identified in each of the following five groups: normal, koilocytosis, low grade squamous intraepithelial lesions (CIN 1), high grade squamous intraepithelial lesions (CIN 2) and high grade squamous intraepithelial lesions (CIN 3). The Crocker technique was used to stain the sections cut 3 microns thick. With ths silver stain the nucleolar organizer regions (NORs) are stained black and referred to as AgNORs. It has been shown that malignant and premalignant changes in cells produce an increase in AgNORs. In each case eight images were captured using a 100x oil-immersion objective and stored in a Datacube Maxvideo system as 512 x 480 pixels in an 8-bit grayscale per image. The images were processed using the NeoPath field-of-view computer to detect the AgNORs and nuclei by using grayscale mathematical morphology algorithms. Color overlays of the AgNORs and nuclei were created using segmentation algorithms. The results show that it is possible to differentiate between low grade squamous intraepithelial lesions (CIN 1) and high grade squamous intraepithelial lesions (CIN 2 and CIN 3) taken together; however, there is no difference between low grade squamous intraepithelial lesions (CIN 1) and koilocytosis. The results support the concept that dysplasia cannot be classified effectively into three grades and that low grade squamous intraepithelial lesions (mild dysplasia [CIN 1]) is indistinguishable from koilocytosis.

Analysis of Variance↗

PubMedDate not supplied

A feasibility study of the AutoPap system location-guided screening.

OBJECTIVE: To study the feasibility of AutoPap System location-guided screening and to evaluate the accuracy of the AutoPap System in selecting locations of potentially abnormal cellular material, the accuracy of triaging slides using the selected locations and the improvement in laboratory accuracy and workload reduction resulting from a combination of location-guided rescreening and AutoPap System primary screening. STUDY DESIGN: For the study, 683 conventionally prepared cervical cytologic smears (263 WNL, 184 atypical squamous cells of undetermined significance/atypical glandular cells of undetermined significance, 173 low grade squamous intraepithelial lesions, 55 high grade squamous intraepithelial lesions and 8 cancer slides) were acquired from nine laboratories. The study slides were independently screened to establish study reference diagnoses. The slides were processed on the AutoPap System and screened by cytotechnologists who were guided by the field of view (FOV) locations provided by the AutoPap System. The location-guided information was used either for initial manual screening (location-guided screening or after initial manual screening of the entire slide (location-guided rescreening). The results were compared to the study reference diagnoses. Sensitivity and workload reduction figures were estimated for different slide populations. RESULTS: The accuracy of location-guided slide triage was high and did not vary significantly over different slide populations. Also, for > 80% of the abnormal slides, the AutoPap System-selected FOV slide locations did contain the most diagnostic cells, which were used to assign an accurate diagnosis to the slide. By combining location-guided rescreening with AutoPap primary screening, the estimated overall location-guided screening false negative fraction (FNF) for LSIL+ slides was approximately 40 times lower than the cytotechnologist-only FNFs, and this improvement was achieved-even at a no review rate of 70%. CONCLUSION: The feasibility study results showed that the AutoPap System location-guided screening process can triage slides accurately and enhance the overall accuracy of slide diagnosis. The combination of location-guided rescreening and AutoPap primary screening could improve the accuracy and workload of the laboratory.

Automation↗

PubMedDate not supplied

AutoPap system performance in screening for low prevalence and small cell abnormalities.

OBJECTIVE: To summarize the design principles of the AutoPap System evaluation score by evaluating slides having a low prevalence of abnormal cells and small cell abnormalities and assessing the evaluation score as a diagnostic tool. STUDY DESIGN: Data from two clinical studies conducted using the AutoPap System and data obtained from the evaluation score training slides were analyzed to demonstrate the effectiveness of the evaluation score. The clinical studies included a prospective, intended-use study involving approximately 13,000 slides and a comprehensive sensitivity study using approximately 1,200 slides from five laboratories. The evaluation score training set consisted of 4,174 slides from 10 laboratories. RESULTS: The robust design of the AutoPap evaluation score was demonstrated by similar detection capabilities and sensitivities to slides having either a low or high prevalence of abnormal cells. No significant difference in performance was detected between the small cell slides and the comparison groups of carcinoma in situ and invasive squamous carcinoma having normal-sized abnormal cells. In addition, the evaluation scores corresponded well to the diagnostic severity of the slides.

Algorithms↗