Health & Lifestyle

Poster Presentation at TechEvince 1.0

Poster Presentation at TechEvince 1.0
of 1
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
   An Enhanced Algorithm for the Quantification of Human Chorionic Gonadotropin  hCG Level in Commerciall Available Home Prenanc Test Kits    Kumar Abhishek 1 , Mrinal Haloi 1 , Sumohana S. Channappayya 2 , Sivaramkrishna Vanjari 2 , Dhananjaya Dendukuri 3 , Swathy Sridharan 3 , Tripurary Choudhary 3 , Paridhi Bhandari 3     , , , , , , , 1 Indian Institute of Technology Guwahati, 2 Indian Institute of Technology Hyderabad, 3 Achira Labs Pvt. Ltd. Bangalore T s s s 1, T.   e wor presentedere s currentyunder revewat te atona onerence onommuncatons 1, anpur.  Home pregnancy kits typically provide a qualitative (binary, i.e. YES/NO) result based on the concentration of human chorionic gonadotropin (hCG) present in urine samples. We present an Home pregnancy test kits have become the default tool for early pregnancy detection. These kits are very affordable and provide fairly accurate results. Typically, the test is performed on urine and is a qualitative test where the presence of hCG is determined by lateral flow immunoassaysthat contain antibodies specific to    l Sandwich immunoassay for β -hCG test was carried out on pilot batches of nitrocellulose chip with positive and negative spiked urine samples.   algorithm that converts this purely qualitative test into a semi-quantitative one by processing digital f ' f the β -subunit of hCG. The appearance of a test line implies that the hCG is present beyond the detection threshold of the assay stripthat is being used. The minimum detection threshold for most of the commercially available stris is around 25 mIUml. Thouh a ualitative test is sufficient to confirm renancy   o  c Analyte ( β -hCG) concentrations in the range from 25 mIU/ml to 1000 mIU/ml were spiked into human male urine samples. Test was performed   . classifies an input into one of four different hCG concentration levels based on the color of the test . , quantification of hCG is necessary to predict different anomalies in pregnancy. For example, a normal pregnancy is indicated by a doubling of the hormone every 48 f f f   o  n    r  o   t  by adding 3 drops (75ul) of reactive sample onto sample pad located upstream of the detection antibody zone of the commercial strip (Cipla, an     a  c ne. e propose agortm proves sgncant mprovement over a pror meto an reuces the maximum false positive rate to less than 5%. This improvement is achieved by a careful choice . . carried out on urine samples and is based on Enzyme Linked Immunosorbent Assay (ELISA) technique. It is costly, time consuming and requires skilled technicians to f -     c   t   i    l   P  , Indian manufacturer). Lateral flow of sample along test strips through caillary action reconstitutes the conuate and form old-labeled     s   t  r of the color space so as to maximize the inter-concentration separability. Also, the proposed method increases the utility of the test kits by providing useful diagnostic information. . - .    d  u   n   t  detection antibody-analyte complex. This labeled immune complex    A Furthermore, the algorithm could be ported to a mobile platform to make it particularly helpful in remote rural health monitoring.ecenty, tere as een growng nterest n usng ow cost magng evces e moe pones to quant   y suc tests at ome[][][]. e propose agortm s an attempt to reduce the cost by using cheap processing power available on mobile devices as opposed to relatively expensive ELISA plate readers or autoanalyzers      t  r   m  e  immobilized in the test and control line producing two red lines visible to in identifying the concentration of β -hCGand is an extension and improvement over the previous work done by ManasaK. et al[1]. The algorithm relies on the acquisition and processing of a digital image of the kit's output. Further, the algorithm was developed using data collected from commercially available off-the-shelf    I   e  r   i  te nae eye, ncatng te successu nng o protens. mages o these tested strips were captured under single bright white light lamp pregnancy detection kits.    E  x  source from a distance of 15 cm by Nikon D5100 DSLR camera fixed at tripod stand.     m Input Image Segmentation Identify Test Color Output      i   t   h  Image Preprocessing to find lines lines Classification ConcentrationRegion of Interest  Image Preprocessing : Inconsistent lighting is a predominant issue during image acquisition. The image is traversed column-wise and pixel values areaveraged to remove irreularities. A median filter is alied alon the rows.    g  o  r Pregnancy Test Kit   . . Segmentation to find lines : A threshold is identified using Otsu’s algorithm and used to binarize the image. However, it was found that just using a thres hold in the RGB (Red, Green Blue color sace ives relatively oor results. In order to imrove erformance we use the HSV (Hue Saturation Valuecolor sace in addition to the semented      A   l After adjusting for lighting irregularities, . , , , image from the RGB color space as [1]. The minima and maxima are of the sum of the Hue and the Saturation values of these images, along a particular row, and areas f   e   d     . Locating Control and Test Lines: It is assumed that the image is taken such that the control line is always to the right of the image. With reference to the algo rithms defined   o  Lines detected as shown in RGB color spaceor segmentaton aove, a regon s taen as an area o nterest t s common to te regons ocate y ot, te an te coor space agortms. Color Classification: The popular SVM- based classification technique is applied to solve the color classification problem. It was empirically determine d that a polynomial     r  o  Test oints for the SVM classifier reresented in the Obtaining the region of interest-reions common to   kernel SVM classifier gives the best results. A much better alternative for training the classifier was found to be the CIE 197 6 L*u*v* color space. The CIE L*u*v* mimics the human visual system and takes into account the Tristimulusvalues of the color, rather than just their intensity values as RGB does.    P  Lines detected as shown in L*u*v* color space Note the better separability of two concentration levels(100 mIU/mL & 250 mIU/mL) in the L*u*v* color space   CIE L*u*v* color space- the RGB and HSV areasmarked black on the horizontal axis    z Based on a the results, we can conclude that there is a ossibility to uantify with a reat deal of accuracy lateral flow [1] K. Manasa, K. V. S. N. L. Manasa Priya, S. Sadhana Reddy, S. S. Channappayya, S.   n     k The proposed algorithm provides a proof-of-concept for converting OutputI I I I I     , assays using sophisticated image processing techniques which canbe orted onto simle imain latforms. Vanjari, D. Dendukuri, S. Swathy, T. Choudhary, and P. Bhandari, “An automated algorithm for the quantification of hcg level in novel fabric -based home pregnancy test kits,” IEEE   s   i  o     W  o  the qualitative process of early pregnancy detection using self test kits into a semi-uantitative one. We believe that uantifyin -   Input mImL mImL mImL mImL0 mIUmL 98.581.4200     . f  Asilomar Conference on Signals, Systems and Computers 2013, Pacific Grove, CA USA , November 2013.     l  u  s    r  e  . hCG levels would be beneficial both to the manufacturer and the IL .. 25 mIU/mL 1.2298.7800     setting with inter sample variability carefully minimized and tested on [2] O. Mudanyali, S. Padmanabhan, S. Dimitrov, I. Navruz, U. Sikora, and A. Ozcan, “Smart   o  n  n  c     F  u   t  .  100 mIU/mL 0095.224.78  I ..   our we-space concentraton eves. n orer to aceve g resouton classification, good training samples are absolutely essential. This in turn [ rapid diagnostics test reader running on a cell-phone for real-time mapping of epidemics,” in Proceedings of the Second ACM Workshop on Mobile Systems, Applications, and Services    C  r  e     &  s part o our uture wor, we propose to ncrease te numer o classes in the classifier to make it more sensitive. Further, to mImL ..  is directly dependent on sensitivity of the commercial strip.While the proposed solution is similar in philosophy to [2][3], there are  for HealthCare , p. 1, ACM, 2012.   s   e   f    i  o  n  aceve our gger goa, te mage processng wou e one entirely on a mobile platform such as Android or iOS.onuson matrx o casscaton resuts or te propose agortm.The classification result is shown as confusion matrix. The first column of the table represents the input two important differences. The images are acquired using a regular digital camera and does not require specialized acquisition devices. [3] O. Mudanyali, S. Dimitrov, U. Sikora, S. Padmanabhan, I. Navruz, and A. Ozcan, “Integrated rapid-diagnostic-test reader platform on a cellphone,” Lab on a Chip , vol. 12, no.   u   l   t    u  s  s  classes and the first row represents the output of the classifier. For example, the second row says that the classifier correctly classified 98.78% of input points belonging to the 25 mIU/ml class and misclassified 1.12%   Further, the proposed algorithm uses a sophisticated SVM-based classifier to provide results that have a resolution higher than binary.  15, pp. 2678–2686, 2012.    R  e     D   i  s  c  of the points as belonging to the 0 mIU/ml class. The classification performance is consistently higher than 95% across classes which is a sinificant imrovement over the reviously obtained 80% in [2].    , .
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks