Archives

  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br In general IoMT devices for

    2020-07-27


    In general, IoMT devices, for example, smartphones and tablets, etc. have low computational power, limited memory, and communication capability. Therefore, most of the IoMT organiza-tions offload all medical records and execute applications on the cloud. IoMT collects health information from patients and then records it to both doctors and clinics through cloud data centers that are further used for diagnosis and clinical services [23,24].  Many IoMT devices generated a large amount of data (big data) which cannot be processed efficiently by conventional data pro-cessing algorithms. Thus, it requires intelligent machine learning algorithms for effective analyses [25,26].
    In a remote situation where there is either lack of medical experts or absence of cancer specialists, the services provided via IoMT can be exploited by providing data in the form of cytol-ogy images through mobile devices to the e-Health care expert system to detect and classify cancer cells. The services provided by the e-Health care server (which can either be based on fog or cloud) is to process and analyze the input data for the detection and classification of breast cells. The main contributions in the proposed approach can be summarized in the following steps:
    • To provide a framework for an e-Health service to remote patients based on IoMT for the analysis of malignant cells.
    • To use both shape and texture based features for accurate classification of malignant N-octanoyl-L-Homoserine lactone in breast cytology images. • To use evolutionary algorithms, i.e., Chain-like Agents in Ge-netics Algorithm (CAGA) for the optimal selection of features and efficient processing of the learning process for accurate classifications.
    • To use ensemble-based classifier for the selection of best classifier by applying the rule of majority voting.
    The rest of the paper is organized as follows: Section 2 pro-vides a detailed analysis of the proposed approach, which in-cludes subsections, i.e., image acquisition and pre-processing, image segmentation, feature computation, and classification. Sim-ilarly, Section 3 discusses the experimental results obtained after applying the proposed approach along with a discussion of its performance evaluation. Finally, Section 4 gives the conclusion of the paper and provides future directions.
    2. Proposed methodology
    Cloud-based e-Health care systems are evolving at a rapid pace due to the advancement in ICT technologies, such as 5G technol-ogy and IOMT, etc. Consider a scenario, where remote e-Health care centers are limited in resources (i.e., lack of medical devices or lack of medical experts) to determine the health condition of a patient who has breast cancer. In this situation, the local health care center transmits patient breast cells cytology images using communication devices, for example, mobile, computer/laptop, or tablet supporting IoMT technologies to the cloud-based e-Health care application server. As the data arrives, using the proposed approach, the cloud-based e-Health care application server can efficiently detect and classify the malignancy of breast cells and transmit back the outcomes to the local health care facility for necessary medical safety actions, as presented in Fig. 1.
    In the cloud-based e-Health care application server, the pro-posed approach is executed which is divided into four phases: image acquisition and pre-processing, image segmentation, fea-tures extraction, and classification using ensemble learning. The data flow diagram of the proposed approach is shown in Fig. 2.
    2.1. Image acquisition and pre-processing
    The breast microscopic images include numerous types of cells and other fluids objects. Due to the copiousness of cells in the image with other objects, it is difficult to detect the cell with-out pre-processing. For this purpose, first linear transformation method, given in [27], is applied to the input image to enhance the contrast between the foreground and background objects. Due to the high correlation in red, green, and blue (RGB) com-ponents in the input image, the processing of each component increases the computational complexity. Thus, conversion of the input image from the RGB color space to the Lab color space is
    Fig. 1. IoMT based e-Health care.
    performed, where L denotes the lightness, and a and b represent the dimension of color [28]. The conversion to Lab color space results in perceptual uniformity. Similarly, it also enhances the contrast of small regions in the image [29].
    Due to linear transformation and conversion to Lab, the whole image is enhanced equally. Thus, the noise contents or unwanted objects in the image are also improved, making it difficult for other image processing techniques to detect and classify malig-nant cells. In image processing applications, filtration is a con-ventional process to smooth the low frequency and enhance the higher frequency for noise removal and feature simplification. There are two common types of filters, i.e., linear and nonlinear, in image processing applications. It is essential to determine the type of filter to be used, which depends on the nature of the image and the application scenario. If the image has less amount of noise but the impact of noise is comparatively high, then a non-linear filter will give good results. In the proposed approach, to remove the single pixel noise, a non-linear median filter is applied which is commonly used for speckle noise and non-linear noise [30]. N-octanoyl-L-Homoserine lactone The essential characteristic of the non-linear filter is such that it does not disturb the image sharpness during the noise removing process. Similarly, the performance result of median filter changes with the size of the mask [13].