Study of Clinical Sample Detection for LH With Lateral Flow Immunochromatographic Strip Using Support Vector Regression.

Abstract

Membrane-based lateral flow immuno-chromatographic strip (LFICS) is widely used in the measurement of Luteinizing hormone (LH) because of its advantages such as easy to handle, low costs, room temperature storage and transport, no need for long storage or transportation of samples and no requirement for skilled technicians. However, LFICS can only provide qualitative or semi-quantitative results without quantitative information, which greatly limited its application. In this paper, we developed a novel quantitative detection method for LFICS using support vector regression (SVR). Canny edge detection operator and fuzzy c-means (FCM) clustering algorithm were also used to extract test line, control line and background part from LFICS images taken by smartphone. The features extracted from test line, control line and background part were used as the input features of SVR model to obtain LH concentration. Measurements of standard sample and clinical sample using proposed method were conducted. Concentration results of standard LH solutions obtained from this method showed a fine linear relationship (r = 0.985) from 1.0 to 250.0 mIU/mL. Seventy-eight clinical serum samples were detected and its corresponding correlation coefficient was 0.918. The method was used to track the urine LH level of a volunteer during ovulation, and quantitative results could be obtained within 15 min.