Frontline clinical diagnosis-FTIR on pancreatic cancer

Abstract

Objective: Accurate, easily accessible and economically viable cancer diagnostic tools are pivotal in improving the abysmal 5% survival rate of pancreatic cancer.

Methods: A novel, affordable, non-invasive diagnostic method has been developed by combining measurement precision of infrared spectroscopy with classification using machine learning tools.

Results: Diagnosis accuracy as high as 90% has been achieved. The study investigated urine and blood from pancreas cancer patients and healthy volunteers, and significantly improved accuracy by focusing on sweet-spots within blood plasma fractions containing molecules within a narrow range of molecular weights.

Keywords: FTIR; PCA; SVM; biomarker; cancer; diagnosis; pancreatic; spectroscopy.

FIGURE 1

Summary of cancer diagnostic techniques. High lighted by circles, Pancreatic cancer work (shown as ■) and oral cancer work (shown as ▲) on plasma are from our research group on 74 pancreatic and 120 oral cancer patients. Raman data are also from our own research on 90 patients' serum, though the 10–30 kDa variation is on an 18‐patient subset. Head/Neck/Brain cancer data from a recent study by Van Der Hoorn et al. using 854 patients for the MRI and 287 for the apparent diffusion coefficient (ADC) imaging metric and clinical validation result from an IR brain cancer study by Butler et al. (shown as ●). Mammogram (MG) combined with clinical examination (CBE) were from a breast cancer (shown as ◆) study on 32,080 patients by Noriaki et al. The dotted curve represents a qualitative boundary of acceptability followed by us with a minimum 85% accuracy. Accuracy is defined as the % of cases classified correctly.

FIGURE 2

Classification accuracies between FTIR spectra of pancreatic premalignant and cancer patients for different plasma and urine molecular weight subsets. Errors are standard binomial error calculations.

FIGURE 3

Difference in the average spectra of cancer and premalignant patient serum from healthy. (A) <10 kDa, (B) whole serum. Faded colour around each line shows the spread for different patients.