Pancreatic tumors are characterized by marked deposition of extra-cellular matrix, also called desmoplasia, which interacts with tumor cells and facilitates the tumor onset and progression. Thus, it would be relevant to develop a method to quantitatively assess the amount of desmoplasia in images derived from bioptic tissue fragments of the pancreas. To this purpose, we applied the principles of fractal geometry, for the assessment of the fractal dimension of images of Masson's trichrome stained pancreatic tissue. Thus, we implemented an algorithm for the computation of the Hausdorff dimension, based on the box counting method: the image is split into boxes of identical size, and the number of boxes needed to cover the features of interest in the image is counted. The process is then iterated with boxes of lower size, and finally all box counts obtained at the different steps are considered, to get the estimate of the Hausdorff dimension, D. After validating the algorithm with appropriate tests, we applied it to pancreatic images, where some regions of interest (ROI) were identified, including both healthy and non-healthy (fibrotic) tissue. We found that non-healthy ROI typically show higher D values than healthy ROI (1.927±0.086 vs. 1.750±0.070 (mean±SD), p=0.0013). Thus, our approach may be of help for an accurate quantification of the degree of severity of pancreatic tumors.
Assessment of the fractal dimension of images derived by biopsy of pancreatic tissue: Implications for tumor diagnosis
Tura A;Sbrignadello S;Finesso L
2016
Abstract
Pancreatic tumors are characterized by marked deposition of extra-cellular matrix, also called desmoplasia, which interacts with tumor cells and facilitates the tumor onset and progression. Thus, it would be relevant to develop a method to quantitatively assess the amount of desmoplasia in images derived from bioptic tissue fragments of the pancreas. To this purpose, we applied the principles of fractal geometry, for the assessment of the fractal dimension of images of Masson's trichrome stained pancreatic tissue. Thus, we implemented an algorithm for the computation of the Hausdorff dimension, based on the box counting method: the image is split into boxes of identical size, and the number of boxes needed to cover the features of interest in the image is counted. The process is then iterated with boxes of lower size, and finally all box counts obtained at the different steps are considered, to get the estimate of the Hausdorff dimension, D. After validating the algorithm with appropriate tests, we applied it to pancreatic images, where some regions of interest (ROI) were identified, including both healthy and non-healthy (fibrotic) tissue. We found that non-healthy ROI typically show higher D values than healthy ROI (1.927±0.086 vs. 1.750±0.070 (mean±SD), p=0.0013). Thus, our approach may be of help for an accurate quantification of the degree of severity of pancreatic tumors.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.