The evaluation value of fuzzy classification of EUS texture features in the therapy of radioactive seeds implatation in pancreatic cancer

Aims: Eus-guided radioactive seeds implantion has been found to be effective that may improve quality of life and survival in patients with advanced unresectable pancreatic cancer. The evaluation of it's effects remains a challenging clinical problem.

Objective: To test the evaluation value of fuzzy classification of ultrasonography (EUS) texture features in radioactive seeds-implanting treatment in patients with pancreatic cancer.

Methods: The method of fuzzy classification in the classification of EUS images was introduced. The seeds treatment effect was evaluated through comparing the probability of cancer about before-treatment and after-treatment cases. 22 texture features are extracted according to the digital image processing algorithm. 216 EUS images are used as the train set and test set to perform several random experiments. Then the other 360 EUS images of 25 patients including before-treatment and after-treatment states are classified fuzzily according to the preceding best training results. Gathering patients' survival time, tumor's volumes and carbohydrate atigen 19–9 (CA19–9). Then the tumor's volumes' change, CA19–9's variation, and variation of probability of cancer (both before-treatment and after-treatment) are comparing to patients' survival time (whether more than 3 months).

Results: 19 patients' variation of probability of cancer vary with the patients' survival time (whether more than 3 months), which is much higher than the rates of volumes' change (14 patients in 25) and CA19–9's variation (9 patients in 25).

Conclusion: The proposed method result in a high classification accuracy and basically follows the change of clinical parameters. so it can to some extent be utilized as a supplementary assessment on the effects of the radioactive particle-implanting treatment in patients with pancreatic cancer.