Taxonomic Classification of Phytoplankton with Multivariate Optical Computing, Part III: Demonstration

Abstract

We describe the automatic analysis of fluorescence tracks of phytoplankton recorded with a fluorescence imaging photometer. The optical components and construction of the photometer were described in Part I and Part II of this series in this issue. An algorithm first isolates tracks corresponding to a single phytoplankter transit in the nominal focal plane of a flow cell. Then, the fluorescence streaks in the track that correspond to individual optical elements on the filter wheel are identified. The fluorescence intensity of each streak is integrated and used to calculate ratios. This approach was tested using 853 fluorescence measurements of the coccolithophore Emiliania huxleyi and the diatom Thalassiosira pseudonana. Average intensity ratios for the two classes closely follow those predicted in Part I of this series, with a distribution of ratios in each class that is consistent with the signal-to-noise ratio calculations in Part II for single cells. No overlap of the two class ratios was observed, yielding perfect classification.

Asymmetric Versus Symmetric Filter Wheels and Associated Processing Algorithms: Results from Asynchronous Fluorescence Imaging Photometer Measurements of Phytoplankton

“…Track Identification. All sample images are first background subtracted and flat-field corrected using the procedure described by Pearl et al 3 Phytoplankton ''tracks'' (areas of images that contain phytoplankton fluorescence) are then identified using autocorrelation of each column to determine whether a track is likely to be present, again, using the procedure described by Pearl et al 3 An autocorrelation threshold is established for each image set by determining the autocorrelation of all columns in a background image set (corrected in the same manner as the sample image set) and setting the threshold as the average plus-two standard deviations of the average background autocorrelation of each column. Adjacent columns that exceed the threshold in each image are grouped together and edges of these groups are identified to give a preliminary estimate of the bounds of potential tracks in the set.…”

“…In past reports, our lab has described the use of a filter wheel on the excitation arm of an asynchronous fluorescence imaging photometer (FIP) for classification of phytoplankton cells. [1][2][3][4] The asynchronous character of the measurements resulted from the fact that phytoplankton entered the field of view of the instrument at random times, so there was no exact correlation between the appearance of a phytoplankton cell and the position of the filter wheel. Instead, the position of the filter wheel had to be inferred from measurements of the phytoplankton fluorescence recorded on a single charge-coupled device (CCD) camera frame as the cell passed through the field of view of the camera.…”

“…One drawback of the original filter wheel was that one opening was blocked to provide a reference timing mark in data from each rotation of the wheel. This allowed us to match fluorescence measurements to filters in the remaining five positions as detailed by Pearl et al 3 Thus, one of the filter wheel positions was sacrificed for timing purposes, and only five data channels were usable, meaning that only two unique matched pairs of filters could be used instead of three. An alternate form of timing could recover the sixth filter channel and enable a third unique pair in the same wheel.…”

Single-Cell and Bulk Fluorescence Excitation Signatures of Seven Phytoplankton Species During Nitrogen Depletion and Resupply

“…12,13 Most of our work has been conducted with phytoplankton cultures grown under non-varying incident irradiance, fixed light-dark cycles, and under nutrient-replete conditions. 8,14 Extension of our results from cultures to natural populations requires an understanding of how environmental fluctuations might affect relative pigment concentrations and hence fluorescence ratio signatures for cells. Cellular pigment ratios are known to change in response to light and nutrient status in at least some phytoplankton taxa.…”

Filter design for molecular factor computing using wavelet functions

“…Nelson et al designed a filter for multivariate optical computing (MOC) based on the loadings from principal component analysis (PCA), and applied it in the quantitative analysis of ethanol in water [6]. Then, the MOC filters based on the loading of linear discriminant analysis (LDA) were designed and used for classification of phytoplankton [7][8][9]. Unlike the molecular filters, MOC filters were made by depositing metallic oxides to a glass substrate [10][11][12][13].…”