Fractal study on nuclear boundary of cancer cells in urinary smears

Cancer is a serious problem for human being and is becoming
a serious problem day-by-day .A prerequisite for any therapeutic modality is early diagnosis.
Automated cancer diagnosis by automatic image feature extraction procedures can be used as
a feature extraction in the field of fractal dimension. The aim of this survey was to introduce a
quantitative and objective mathematical method for pinpointing the differences between malignant
and non –malignant epithelial cells in urine cytology by the use of software analysis.
Materials & Methods: Forty-one positive urine cytology and 33 negative subjects from
Pathology Department of Imam Khomeini Hospital, Urmia, Iran (2003-2007) were selected at
random. Digitalized images were prepared by the use of objective 100X (a digital video head) which
subsequently were processed by the BeonitTM software version 1.3 (Tru Soft International inc. USA)
to measure fractal dimension of nuclear boundaries.
Results: Findings revealed statistically significant differences between fractal dimensions of
nuclear boundaries of cancerous and non-cancerous smears (P=0.001). Study had selected a cut-off
point to (1.732 ± 0.006) to discriminate malignant and non-malignant epithelial cells in urinary smears.
Conclusion: Based on diagnostic accuracy measures (sensitivity and specificity), probability of
disease measures (predictive value of a positive and negative test results), and likelihood ratio of
positive and negative tests, it seems fractal dimension of nuclear cell boundaries for urinary smears
can be used as a feature extraction in the field of automated cancer diagnosis