Журнал цитологии и гистологии

Computer Assisted Pap Smear Analyser for Cervical Cancer Screening using Quantitative Microscopy

Abstract

Rajasekharan Usha Deepak, Ramakrishnan Rajesh Kumar, Neendoorthalackal Balakrishnan Byju1 Pundluvalu Nataraju Sharathkumar, Chandran Pournami, Salam Sibi, Ewert Bengtsson and Kunjuraman Sujathan

Cervical cancer is the third most common cancer among women. The bulk of the cancer burden is on low and middle income countries where screening is mostly opportunistic rather than systematic. Among the number of screening methods, cytology based screening using Pap smear test is by far the most widely followed and accepted method. In countries where organized screening using Pap test has been introduced, incidence and mortality caused by the disease has significantly subsided. Although the method is effective in controlling the disease, it poses a serious challenge in practical implementation owing to the fact that the method is resource intensive requiring trained professionals skilled enough to identify a handful of abnormal cells among few hundred thousand cells. This motivates the need for automating the screening methodology. Since the 1960-ies numerous projects have developed such automated screening systems leading also to a couple of commercial products. Still these have had limited impact on the screening situation in most of the world. This paper describes a screening system developed by our group in an effort of creating a cost effective screening system that could be widely deployed. The systems digitizes Pap smear slides and carries out cell level and smear level analysis on digitized smear and finally classifies the smear as either normal or suspicious. Clearly normal smears were screened out without any human intervention while suspicious smears were sent for expert cytologist review. A low cost monolayer slide preparation technique has also been identified which produces monolayer slides of quality comparable to that of commercial systems at much lesser cost. The computer aided Pap smear analyzer was validated at the Regional Cancer Centre (RCC), Thiruvananthapuram, India since May 2011. Since then a total of 1107 smears covering all abnormal and normal categories has been evaluated with a specificity of 60% and overall sensitivity of 80%. The system produces even higher sensitivity of 93% and 95% in HSIL and SCC grades respectively. Each slide used for validation has undergone two arm blind reviews, first by conventional manual cytology by qualified cytologists and second by automated Pap smear analysis. The accuracy of the automated analysis was benchmarked by using the manual review result as gold standard. The system has been found to reduce the workload of cytologist to almost 60% and has been designed to be operated by a semi-skilled person. A fully automated system can be builtbased on the results obtained by the present system by adding a slide loader, scanner, bar-code reader, sensors etc. which when designed and build cost effectively can increase slide processing throughput and reduce the dependency on human labour thereby significantly reducing the cost per slide making it feasible to extend screening to many more.