Vol.15, No 1, 2008 pp. 28 - 32
UC 616.24-006-073
DISCRIMINANT ANALYSIS OF NUCLEAR IMAGE VARIABLES IN LUNG CARCINOMA
Žaklina Mijović1, Dragan Mihailović1, Miloš Kostov2
1Medical Faculty, Institute of Pathology, Niš
2Military Hospital, Department of Pathology, Niš
 E-mail: zami@bankerinter.net
Summary. Histological classification of lung cancer is subjective and often difficult to reproduce on bronchoscopic biopsies. The aim of this study was to identify which of seven karyometric variables are of diagnostic value in distinguishing major histological types of lung carcinoma. Bronchoscopic mucosal samples from patients with squamous cell carcinoma (n=48), small cell carcinoma (n=41), and adenocarcinoma (n=35) of the lung were retrieved. Specimens were stained with hematoxylin and eosin, and measured by image analyzer. All measured nuclear variables were found to be significantly different between small cell carcinoma and non-small cell carcinoma (p<0.01). Using stepwise discriminant function analysis, a correct diagnosis was achieved in 92.7% of non-small cell carcinomas and 90.24% of small cell carcinomas. The total percent of correct classification was 91.93%. The best discriminant variables between these categories of lung carcinoma were nuclear perimeter and area. In conclusion, nuclear image analysis can be used to make distinction between major histological types of lung carcinoma in the biopsy specimens, especially between small cell lung carcinoma and non-small cell lung carcinoma.
Key words: Computer-assisted diagnosis; image analysis; lung carcinoma
DISKRIMINANTNA ANALIZA NUKLEARNIH IMIDŽ VARIJABLI KARCINOMA PLUĆA
Kratak sadržaj: Histološka klasifikacija karcinoma pluća je subjektivna i često teško reproducibilna na bronhoskopskim biopsijama. Cilj rada bio je da se utvrdi koja od sedam kariometrijskih varijabli ima dijagnostički značaj u razdvajanju glavnih histoloških tipova karcinoma pluća. Izdvojene su bronhoskopske mukozne biopsije pacijenata sa planocelularnim karcinomom (n=48), mikrocelularnim karcinomom (n=41) i adenokarcinomom pluća (n=35). Uzorci su bojeni hematoksilinom i eozinom i mereni imidž analizatorom. Za sve merene nuklearne varijable nađene su signifikantne razlike između sitnoćelijskog i ne-sitnoćelijskog karcinoma pluća (p<0.01). “Stepwise” discriminantnom funkcionom analizom je korektna dijagnoza postignuta u 92,7% nesitnoćelijskih karcinoma i 90,24% sitnoćelijskih karcinoma. Ukupni procenat korektne klasifikacije je 91,93%. Najbolje diskriminantne varijable između ovih kategorija karcinoma pluća su nuklearni perimetar i areal. U zaključku, nuklearna imidž analiza se može koristiti za razlikovanje glavnih histoloških tipova karcinoma pluća na biopsijskim uzorcima, naročito sitnoćelijskog i ne-sitnoćelijskog karcinoma.
Ključne reči: Kompjuterski asistirana dijagnoza, imidž analiza, karcinom pluća