L-CNN: Exploiting labeling latency in a CNN learning framework

Muhammad Jamal Afridi, Arun Ross, Erik M. Shapiro

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    A supervised learning system requires labeled data during the training phase. Obtaining labels can be an expensive process, especially in medical imaging applications where a qualified expert may be needed to carefully analyze images and annotate them. This constrains the amount of labeled data available. This study explores the possibility of incorporating labeling behavior (viz., labeling latency) in a supervised convolutional neural network (CNN) framework in order to improve its performance in the presence of limited labeled data. The problem of 'spot' detection in MRI scans is considered in this work. In this two-class problem, (a) labeling behavior is available only during the training phase unlike traditional features that are available both during training and testing; and (b) the labeling behavior is associated with only one class (the positive samples) unlike other side information that is available for all classes. To address these issues, a new CNN architecture referred to as L-CNN is designed. The proposed method utilizes the labeling behavior of the expert to cluster the labeled data into multiple categories; a source CNN is then trained to distinguish between these categories. Next, a transfer learning paradigm is used where a target CNN is initialized using this source CNN and its weights updated with the limited labeled data that is available. Experimental results on an existing MRI database show that the proposed L-CNN performs better than a conventional CNN and, further, significantly outperforms the previous state-of-the-art, thereby establishing a new baseline for 'spot' detection in MRI.

    Original languageEnglish (US)
    Title of host publication2016 23rd International Conference on Pattern Recognition, ICPR 2016
    PublisherInstitute of Electrical and Electronics Engineers Inc.
    Pages2156-2161
    Number of pages6
    ISBN (Electronic)9781509048472
    DOIs
    StatePublished - Apr 13 2017
    Event23rd International Conference on Pattern Recognition, ICPR 2016 - Cancun, Mexico

    Other

    Other23rd International Conference on Pattern Recognition, ICPR 2016
    CountryMexico
    CityCancun
    Period12/4/1612/8/16

    Profile

    Neural networks
    Labeling
    Magnetic resonance imaging
    Supervised learning
    Medical imaging
    Network architecture
    Learning systems
    Labels
    Testing

    ASJC Scopus subject areas

    • Computer Vision and Pattern Recognition

    Cite this

    Afridi, M. J., Ross, A., & Shapiro, E. M. (2017). L-CNN: Exploiting labeling latency in a CNN learning framework. In 2016 23rd International Conference on Pattern Recognition, ICPR 2016 (pp. 2156-2161). [7899955] Institute of Electrical and Electronics Engineers Inc.. DOI: 10.1109/ICPR.2016.7899955

    L-CNN : Exploiting labeling latency in a CNN learning framework. / Afridi, Muhammad Jamal; Ross, Arun; Shapiro, Erik M.

    2016 23rd International Conference on Pattern Recognition, ICPR 2016. Institute of Electrical and Electronics Engineers Inc., 2017. p. 2156-2161 7899955.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Afridi, MJ, Ross, A & Shapiro, EM 2017, L-CNN: Exploiting labeling latency in a CNN learning framework. in 2016 23rd International Conference on Pattern Recognition, ICPR 2016., 7899955, Institute of Electrical and Electronics Engineers Inc., pp. 2156-2161, 23rd International Conference on Pattern Recognition, ICPR 2016, Cancun, Mexico, 4-8 December. DOI: 10.1109/ICPR.2016.7899955
    Afridi MJ, Ross A, Shapiro EM. L-CNN: Exploiting labeling latency in a CNN learning framework. In 2016 23rd International Conference on Pattern Recognition, ICPR 2016. Institute of Electrical and Electronics Engineers Inc.2017. p. 2156-2161. 7899955. Available from, DOI: 10.1109/ICPR.2016.7899955

    Afridi, Muhammad Jamal; Ross, Arun; Shapiro, Erik M. / L-CNN : Exploiting labeling latency in a CNN learning framework.

    2016 23rd International Conference on Pattern Recognition, ICPR 2016. Institute of Electrical and Electronics Engineers Inc., 2017. p. 2156-2161 7899955.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    @inbook{f99519e6b9ff44e1a44a93ddc3e77832,
    title = "L-CNN: Exploiting labeling latency in a CNN learning framework",
    abstract = "A supervised learning system requires labeled data during the training phase. Obtaining labels can be an expensive process, especially in medical imaging applications where a qualified expert may be needed to carefully analyze images and annotate them. This constrains the amount of labeled data available. This study explores the possibility of incorporating labeling behavior (viz., labeling latency) in a supervised convolutional neural network (CNN) framework in order to improve its performance in the presence of limited labeled data. The problem of 'spot' detection in MRI scans is considered in this work. In this two-class problem, (a) labeling behavior is available only during the training phase unlike traditional features that are available both during training and testing; and (b) the labeling behavior is associated with only one class (the positive samples) unlike other side information that is available for all classes. To address these issues, a new CNN architecture referred to as L-CNN is designed. The proposed method utilizes the labeling behavior of the expert to cluster the labeled data into multiple categories; a source CNN is then trained to distinguish between these categories. Next, a transfer learning paradigm is used where a target CNN is initialized using this source CNN and its weights updated with the limited labeled data that is available. Experimental results on an existing MRI database show that the proposed L-CNN performs better than a conventional CNN and, further, significantly outperforms the previous state-of-the-art, thereby establishing a new baseline for 'spot' detection in MRI.",
    author = "Afridi, {Muhammad Jamal} and Arun Ross and Shapiro, {Erik M.}",
    year = "2017",
    month = "4",
    doi = "10.1109/ICPR.2016.7899955",
    pages = "2156--2161",
    booktitle = "2016 23rd International Conference on Pattern Recognition, ICPR 2016",
    publisher = "Institute of Electrical and Electronics Engineers Inc.",

    }

    TY - CHAP

    T1 - L-CNN

    T2 - Exploiting labeling latency in a CNN learning framework

    AU - Afridi,Muhammad Jamal

    AU - Ross,Arun

    AU - Shapiro,Erik M.

    PY - 2017/4/13

    Y1 - 2017/4/13

    N2 - A supervised learning system requires labeled data during the training phase. Obtaining labels can be an expensive process, especially in medical imaging applications where a qualified expert may be needed to carefully analyze images and annotate them. This constrains the amount of labeled data available. This study explores the possibility of incorporating labeling behavior (viz., labeling latency) in a supervised convolutional neural network (CNN) framework in order to improve its performance in the presence of limited labeled data. The problem of 'spot' detection in MRI scans is considered in this work. In this two-class problem, (a) labeling behavior is available only during the training phase unlike traditional features that are available both during training and testing; and (b) the labeling behavior is associated with only one class (the positive samples) unlike other side information that is available for all classes. To address these issues, a new CNN architecture referred to as L-CNN is designed. The proposed method utilizes the labeling behavior of the expert to cluster the labeled data into multiple categories; a source CNN is then trained to distinguish between these categories. Next, a transfer learning paradigm is used where a target CNN is initialized using this source CNN and its weights updated with the limited labeled data that is available. Experimental results on an existing MRI database show that the proposed L-CNN performs better than a conventional CNN and, further, significantly outperforms the previous state-of-the-art, thereby establishing a new baseline for 'spot' detection in MRI.

    AB - A supervised learning system requires labeled data during the training phase. Obtaining labels can be an expensive process, especially in medical imaging applications where a qualified expert may be needed to carefully analyze images and annotate them. This constrains the amount of labeled data available. This study explores the possibility of incorporating labeling behavior (viz., labeling latency) in a supervised convolutional neural network (CNN) framework in order to improve its performance in the presence of limited labeled data. The problem of 'spot' detection in MRI scans is considered in this work. In this two-class problem, (a) labeling behavior is available only during the training phase unlike traditional features that are available both during training and testing; and (b) the labeling behavior is associated with only one class (the positive samples) unlike other side information that is available for all classes. To address these issues, a new CNN architecture referred to as L-CNN is designed. The proposed method utilizes the labeling behavior of the expert to cluster the labeled data into multiple categories; a source CNN is then trained to distinguish between these categories. Next, a transfer learning paradigm is used where a target CNN is initialized using this source CNN and its weights updated with the limited labeled data that is available. Experimental results on an existing MRI database show that the proposed L-CNN performs better than a conventional CNN and, further, significantly outperforms the previous state-of-the-art, thereby establishing a new baseline for 'spot' detection in MRI.

    UR - http://www.scopus.com/inward/record.url?scp=85019078241&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=85019078241&partnerID=8YFLogxK

    U2 - 10.1109/ICPR.2016.7899955

    DO - 10.1109/ICPR.2016.7899955

    M3 - Conference contribution

    SP - 2156

    EP - 2161

    BT - 2016 23rd International Conference on Pattern Recognition, ICPR 2016

    PB - Institute of Electrical and Electronics Engineers Inc.

    ER -