Mathematical models of differential diagnostics and prognosis in chronic pancreatitis and cancer with a primary lesion of the pancreatic head
Keywordschronic pancreatitis, cancer of the pancreatic head, differential diagnosis, mathematical models
AbstractBackground. Differential diagnostics of chronic pancreatitis and the pancreas cancer with prevalent affection of head is on of the most difficult and important problems of surgical pancreatology.
Methods. Analyzed the results of surgical treatment of 132 patients, including 68 - for cancer of the pancreatic head (in 46 - with jaundice) and 64 - chronic pancreatitis (CP) with a primary lesion of the pancreatic head (16 - with jaundice). The distribution of patients into groups was carried out with a maximum value of classification functions calculated by special formulas. Next studied indicators of endothelial dysfunction for differential diagnosis.
Results. It was defined the threshold of VEGF = 346 pg/ml, which shared the group of chronic pancreatitis or cancer of the pancreatic head, which was determined based on the Pareto criterion. This model sensitivity was 72.1% and specificity of 75% for the overall accuracy of 72.7%. Even more precision indicator was on the threshold of VEGF = 248 pg/ml, which compared groups of patients with cancer and software of the control group (125.9 pg/ml) and the sensitivity was 86.8%, specificity 82.4%, and overall accuracy of 82.3%. At about the same accuracy had this test and the comparison group of patients with chronic pancreatitis and control: sensitivity 84.4% and specificity of 76.5% overall accuracy of 81.5% in the threshold VEGF of 155 pg/ml (p<0,05). To develop a prognosis of a pathological process, along with the use of diagnostic data used a method of classification trees. The model showed that the index VEGF is the criterion that discriminates for pancreas- pancreatic cancer-pancreas, but relative differences in the presence of jaundice in patients defined using S-nitrozothiol. The accuracy of the proposed method of prediction was 89%, the price of cross-checking - 82,6% (p<0,05). Pancreatoduodenal resection for Whipple was performed in 23 patients, for Traverso-Longmire - in 8, subtotal right sided pancreatectomy for Fortner - in 3, hepaticojejunostomy by Roux - in 8, duodenopreserving resection for Beger - in 6, her Bernese option - in 7, operation Frey - in 51. In 26 (19.7%) patients, minimally invasive intervention for removal of bile were spread through the final primary pathological process and severe general state. Postoperative complications occurred in 18 (13.6%) patients, died 3 (2.3%).
Conclusion. Constructed mathematical models of differential diagnosis and prognosis allow with the high accuracy detect cancer in the presence of the volumetric formation on the head of pancreas. Along with the using of clinical laboratory and instrumental data to decide the issues of the head’s pancreas cancer and chronic pancreatitis diagnostic can be used the proposed method for constructing the classification trees, whose accuracy was 89%, the price of cross-checking - 82.6%, which can be considered as good indexes of the model.
How to Cite
The periodical offers access to content in the Open Access system under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0
Number of views and downloads: 105
Number of citations: 0