Machine learning based detection of pancreatic tumors using the Dxcover cancer platform.

Abstract

e16275 Background: Pancreatic cancer is the 7th most deadly cancer worldwide with over 460,000 victims per year. In the current diagnostic pathway, carbohydrate antigen (CA) 19-9 serum test is the blood assessment used for detection of pancreatic cancer; although, with poor positive predictive values reported, it is not specific for pancreatic tumors as its levels can be raised in symptomatic patients with other benign comorbidities and/or because of other tumors in the surrounding area. Attenuated total reflection–Fourier transform infrared spectroscopy (ATR-FTIR) has demonstrated exceptional potential in human blood serum analysis for cancer diagnostics and its implementation in the clinical environment could represent a significant step forward in the early detection of pancreatic cancer. This proof-of-concept study aimed to investigate the use of the Dxcover cancer liquid biopsy as a novel approach for pancreatic cancer detection. Methods: The study was focused on the discrimination between both cancer (n = 100) versus healthy control samples (n = 100), and cancer (n = 35) versus symptomatic non-malignant control samples (n = 35) from patients with comorbidities and/or confounding diseases. Various machine learning algorithms were applied to discriminate between the classes: random forest (RF), partial least squares-discriminant analysis (PLS-DA) and support vector machine (SVM). Receiver operating characteristic (ROC) analysis was also employed to evaluate the classes’ degree of diagnostic separability. Permutation tests were performed for each classification in order to ensure their statistical significance. Results: Cancer (n = 100) and healthy control samples (n = 100) were distinguished with excellent results, achieving results up to a sensitivity of 91.0 %, specificity of 87.6 %, and accuracy of 89.3 % with PLS-DA. Moreover, an area under the curve (AUC) equal to 0.954 was obtained through ROC analysis. When discriminating between cancer (n = 35) and symptomatic control samples (n = 35), accuracy values were recorded in the range between 70.6 and 77.3 % with ROC analysis showing a balanced sensitivity and specificity over 75 % with an AUC of 0.844. Both discriminations were proven statistically significant. Conclusions: Pancreatic cancer detection in early stages would be key in improving prognosis and survival rates of patients, through targeted earlier surgery and treatments. The Dxcover cancer liquid biopsy could represent a powerful tool in the clinical environment as an easy-to-use, minimally invasive and reliable spectroscopic blood test for detection of pancreatic cancer.