An Organelle Correlation-Guided Feature Selection Approach for Classifying Multi-Label Subcellular Bio-Images

Abstract

Nowadays, with the advances in microscopic imaging, accurate classification of bioimage-based protein subcellular location pattern has attracted as much attention as ever. One of the basic challenging problems is how to select the useful feature components among thousands of potential features to describe the images. This is not an easy task especially considering there is a high ratio of multi-location proteins. Existing feature selection methods seldom take the correlation among different cellular compartments into consideration, and thus may miss some features that will be co-important for several subcellular locations. To deal with this problem, we make use of the important structural correlation among different cellular compartments and propose an organelle structural correlation regularized feature selection method CSF (Common-Sets of Features) in this paper. We formulate the multi-label classification problem by adopting a group-sparsity regularizer to select common subsets of relevant features from different cellular compartments. In addition, we also add a cell structural correlation regularized Laplacian term, which utilizes the prior biological structural information to capture the intrinsic dependency among different cellular compartments. The CSF provides a new feature selection strategy for multi-label bio-image subcellular pattern classifications, and the experimental results also show its superiority when comparing with several existing algorithms.

MIC_Locator: a novel image-based protein subcellular location multi-label prediction model based on multi-scale monogenic signal representation and intensity encoding strategy

“…Hence, a carefully checking task has been carried out on the corresponding benchmark dataset of two published papers [16, 17]. These benchmark datasets derive from published literature in [16, 17], which are respectively single-label dataset and multi-label dataset and has been used in references [15, 40]. The benchmark datasets in [16] based on the early version of HPA database, and the other benchmark datasets proposed by the Xu et al [17] are collected from the 12 version of HPA database.…”

“…CLBP adds coding the property of center pixels on the basis of LBP. The Haralick texture and DNA spatial distribution feature are one of the most discriminative features of SLFs to describe the IHC image from a global perspective, and it has been widely used in many works and has validated its high-performance [15–17, 31, 34, 40, 41]. In this paper, the SLFs feature, derived from the combination of Haralick feature and the DNA distribution feature, is unified into global feature in total 840-dimension [54].…”

“…Moreover, many efforts have been made on the algorithm level [39–41]. Wei group proposed a novel feature selection method that used the biology background to set up a regularization item so as to optimize the feature selection method, and this method can select more informative feature subsets [40]. The Sullivan group innovatively used the online game (EVE Online) to attract the numerous participants to annotate the subcellular locations of protein image based on both of the transfer learning framework and the deep learning method to build the automated Localization Cellular Annotation Tool (Loc-CAT).…”

PScL-2LSAESM: bioimage-based prediction of protein subcellular localization by integrating heterogeneous features with the two-level SAE-SM and mean ensemble method

“…Current studies have shown that extracting both global and local features can help improve the predictive capabilities of the developed methods ( Xu et al , 2013 , 2016 ; Yang et al , 2019 ). Similarly, at the FS stage, a number of studies have proposed different FS algorithms to effectively select the optimal features from the extracted features ( Liu et al , 2020 ; Newberg and Murphy, 2008 ; Shao et al , 2018 ; Ullah et al , 2021 ). Among such FS algorithms, the stepwise discriminant analysis (SDA) algorithm ( Klecka, 1980 ) has been widely adopted and shown to be effective for FS.…”

Lungs Cancer Detection Using Digital Image Processing Techniques: A Review

“…For example, we found that tumor has intensity values 115-255 then we can convert all values which are less than 115 into 0 and all other values between 115-255 is equal to 1 in binary image. In this way all unwanted area become black and wanted area become white region[13].…”