| Abstract | Cilj istraživanja: Primjenom fraktalne analize na digitaliziranim rendgenskim prikazima kralješka utvrditi mogućnosti razlikovanja zdravog, osteoporotski i tumorski promijenjenog kralješka. Nacrt studije: Istraživanje je ustrojeno kao prospektivno, opservacijsko, presječno istraživanje u kojem su analizirani podatci dobiveni dijagnostičkim pretragama. Bolesnici i metode: U studiju je uključeno 114 pacijenata. Eksperimentalnu skupinu činila su 94 bolesnika s prijelomom kralješaka uzrokovanih osteoporozom odnosno malignim tumorom. Kontrolnu skupinu činilo je 20 bolesnika mlađih od 50 godina koji su liječeni zbog akutnih bolnih križa. Analizirane su digitalizirane rendgenske slike procijenila dva neovisna pregledavatelja. U programu ImageJ u središnjim dijelovima trupova kralješaka označilo se područje interesa (ROI), a programom BoneJ prikupljeni su podatci o fraktalnoj dimenziji, broju koštanih trabekula, broju koštanih trabekula po mm³ te srednja i maksimalna debljina trabekula. Rezultati: Fraktalnom analizom utvrđeno je da ispitanici eksperimentalne skupine imaju značajno manji broj koštanih trabekula te značajno manju vrijednost fraktalne dimenzije u odnosu na kontrolnu skupinu. Kada su se usporedile sve tri skupine (kontrolna, osteoporoza, tumor), značajno su bile više vrijednosti broja koštanih trabekula i fraktalne dimenzije u kontrolnoj skupini u odnosu na ispitanike s osteoporozom. Između vrijednosti kontrolne skupine i skupine s tumorom, kao i skupine s osteoporozom prema onima s tumorom, nije bilo značajnih razlika u promatranim parametrima. Ocijenjena je vjerojatnost prijeloma zbog osteoporoze ili tumora. Univarijantna analiza pokazala je da može razlikovati ispitanike eksperimentalne i kontrolne skupine po kliničkoj slici i točno klasificirati 82,5% slučajeva. ROC analizom utvrđeno je da jedino vrijednost fraktalne dimenzije možemo prikazati kao dijagnostički pokazatelj uz senzitivnost 59,2 i specifičnost 65 te točku razlikovanja ≤ 1,49. Fraktalnom analizom uzoraka eksperimentalne skupine u odnosu na ispitanike s osteporozom ili tumorom nisu pronađene značajne razlike između analiziranih parametara. Zaključak: U otkrivanju patologije analiziranih kralješaka više je testiranih pokazatelja pokazalo određenu vrijednost. Vrijedne informacije dobivene su iz vrijednosti fraktalne dimenzije, dok su isti pokazatelji bili neučinkoviti u razlikovanju uzroka bolesti. Navedeno je u suglasju s više izvora u literaturi i nedvojbeno upućuje na to da patološke promjene u strukturi mijenjaju rendgenski prikaz kralješaka koje je moguće otkriti matematičkom analizom. |
| Abstract (english) | Objectives: Using fractal analysis of digitized vertebral X-ray images to determine the ability to discern healthy, osteoporotic, and tumor-modified vertebrae. Study Design: The study is structured as a prospective, observational, cross-sectional exploration where data analyzed were obtained by diagnostic examinations. Patients and Methods: A prospective study involved 114 patients. An experimental group was comprised of 94 patients treated for vertebral fractures caused by osteoporosis or malignant tumor. The control group was 20 patients younger than 50 who were treated for acute back pain. Digitized x-ray images were analyzed by two independent viewers. In the ImageJ program, the region of interest (ROI) was marked in the central parts of the vertebrae, and the BoneJ program collected data on the fractal dimension, the number of bone trabecules, the number of bone trabecules per mm³, and the mean and maximum thickness of the trabeculae. Results: Fractal analysis showed that experimental groups had significantly smaller bone trabecula and significantly lower fractal dimensions than controls. No significant differences in the number of bone trabecules per mm3, mean and maximum trabecular thickness were found. When compared to all three groups (control, osteoporosis, tumor), bone trabeculae and fractal dimension values were significantly higher in the control group than those with osteoporosis. Comparing values of the control group and the tumor group, as well as the osteoporosis groups with those with the tumor, there were no significant differences in observed parameters. Logistic regression evaluated the influence of analyzed parameters on the likelihood that osteoporosis or tumor fractures would occur in observed subjects. By univariate analysis, significant predictors were maximum thickness of trabecula and fractal dimensions. The model was entirely statistically significant and showed that it could distinguish between experimental and control groups in the clinical picture and accurately classifies 82.5% of the cases. Multivariate regression analysis showed that it can distinguish between experimental and control subjects in a clinical picture, and only the fractal dimension has given a unique statistically significant contribution to the model. ROC analysis showed that, given the presence of osteoporosis or tumor, only the fractal dimension value can be shown as a diagnostic indicator (sensitivity = 59.2, specificity = 65) with a resolution point ≤ 1.49. Fractal analysis of experimental group samples compared to osteoporosis or tumor subjects showed no significant differences between the analyzed parameters. Conclusion: In the discovery of the pathology of the vertebral cortex, a number of tested indicators showed a certain value. Valuable information was obtained from the value of the fractal dimension, while the same indicators were ineffective in determining the cause of the disease. Many sources in the literature agree and beyond a doubt suggests that pathological changes in vertebral structure, which are detected by X-ray, can be demonstrated by mathematical analysis. |
