Urine strip analyzer using artificial neural network through android phone

• Main Author: Velasco, Jessica S.
• Format: text
• Language: English
• Published: Animo Repository 2016
• Subjects: Urine—Analysis; Neural networks (Computer science); Electrical and Computer Engineering; Electrical and Electronics
• Online Access: https://animorepository.dlsu.edu.ph/etd_masteral/6987
• Institution: De La Salle University

Point of Care Testing (POCT) improves clinical process outcome. It has the potential to reduce errors and the wastage of resources. There is a significant amount of information obtained through the examination of urine. The routine urinalysis consists of two major components: physiochemical determination and microscopic examination of urine sediment. The physiochemical determination includes the appearance, specific gravity and reagent strip measurements. The physiochemical properties of urine may include the following analytes: pH, protein, glucose, ketone, blood, biliburin, urobilinogen, nitrite, leukocytes and specific gravity. Reagent strips provide a simple, rapid means for performing medically significant chemical analysis for urine. Assessment of the dipstick test result is done manually by visually comparing the reactive color of each reagent with dipstick color chart based on the color similarities. The manual interpretation has its weaknesses or failure. It includes the differences in a perception of color, differences in lighting condition and a failure to read several reagents in a specified time. The study of artificial neural networks is motivated by its similarity to work with biological systems successfully. It can learn from training samples or by means of neural network capable to learn. After successful training, a neural network can find reasonable solutions for similar problems of the same class that were not explicitly trained. This in turn results in a high degree of fault tolerance against noisy input data. The study developed a urine analyzer in android environment. It is able to read a 4 parameter and 10 parameter urine strip in real-time. This study also used digital image processing that includes cropping, image segmentation, thresholding, smoothing and recognition. The training is different for each parameter. This is done through Levenberg Marquardt. It performed evaluation through comparison of the standard urinalysis and the device. The prototype is evaluated and certified by a professional registered medical technologist. The accuracy test performed proved to have an accuracy of 96%.