Journal of Scientific and Industrial Research

Open Access Open Access  Restricted Access Subscription Access

Rank Based Two Stage Semi-Supervised Deep Learning Model for X-Ray Images Classification


Affiliations
1 Madan Mohan Malaviya University of Technology, Gorakhpur 273 016, Uttar Pradesh, India
2 Galgotias University, Greater Noida 203 201, Uttar Pradesh, India
3 GL Bajaj Institute of Technology and Management, Greater Noida203 201, Uttar Pradesh, India
 

Deep learning approaches rely on a wide-scale labeled dataset to attain a high level of performance. Although labeled data is more difficult and costly to access in some applications, such as bioinformatics and medical imaging, wide variety of ongoing research on the topic of Semi-Supervised Deep Learning (SSDL) can improve and fix underlying problems in this domain. The motivation for the suggested model Rank Based Two-Stage Semi-Supervised Deep Learning (RTS-SS-DL) is the same as how doctors deal with unobserved or suspect cases in day to day practice. The physicians deal with these suspect instances with the help of professional assistance from their colleagues. Before beginning therapy, some patients seek the opinion of a variety of skilled professionals. The patients are treated by the most appropriate (vote count) professional diagnosis. Our model (RTS-SS-DL) has achieved impressive metrics including 92.776% accuracy, 97.376% specificity, 86.932% sensitivity, 96.192% precision, 85.644% MCC (Matthews Correlation Coefficient), 3.808% FDR (False Discovery Rate), 2.624% FPR (False Positive Rate), 91.072% f1-score, 90.85% NPV (Negative Predictive Value), and 13.068% FNR (False Negative Rate) for the unseen dataset. The outcome of this research results in an SSDL model that is both more precise and effective.

Keywords

Labeled dataset, RTS-SS-DL, Self-organising classifier, Semi-supervised learning, Shoulder’s fracture classification.
User
Notifications
Font Size

  • Agrawala A, Learning with a probabilistic teacher, IEEE Trans Inf Theory, 16(4) (1970) 373–379, https://doi.org/10.1109/TIT.1970.1054472.

  • Fralick S, Learning to recognize patterns without a teacher,IEEE Trans Inf Theory,13(1) (1967) 57–64, https://doi.org/10.1109/TIT.1967.1053952.

  • Scudder H, Probability of error of some adaptive patternrecognition machines, IEEE Trans Inf Theory,11(3) (1965) 363–371,https://doi.org/10.1109/TIT.1965.1053799.

  • Rajpurkar P, Irvin J, Bagul A, Ding D, Duan T, Mehta H, Yang B, Zhu K, Laird D, Ball R L & Langlotz C, MURA: Large dataset for abnormality detection in musculoskeletal radiographs,arXiv Prepr arXiv171206957, (2017).

  • Narayan V & Daniel A K, CHHP: coverage optimization and hole healing protocol using sleep and wake-up concept for wireless sensor network, Int J Syst Assur Eng Manag, 13(1) (2022) 546–556.

  • Narayan V & Daniel A K, IOT based sensor monitoring system for smart complex and shopping malls, in Int Conf Mobile Netw Manag (Cham: Springer International Publishing) 2021, 344–354.

  • Narayan V & Daniel A K, A novel approach for cluster head selection using trust function in WSN, Scalable Comput Pract Exp, 22(1) (2021) 1–13.

  • Laine S & Aila T, Temporal ensembling for semi-supervised learning, arXiv Prepr arXiv161002242, (2016).

  • Narayan V & Daniel A K, Multi-tier cluster based smart farming using wireless sensor network, in 2020 5th Int Conf Comput, Commun Secur (IEEE) 2020, 1–5, https://doi.org/ 10.1109/ICCCS49678.2020.9277072.

  • Awasthi S, Srivastava A P, Srivastava S & Narayan V, A Comparative study of various CAPTCHA methods for securing web pages, in 2019 Int Conf Automat Comput Technol Manag (IEEE) 2019, 217–223, https://doi.org/ 10.1109/ICACTM.2019.8776832.

  • Narayan V & Daniel A K, Design consideration and issues in wireless sensor network deployment, Invertis J Sci & Technol, (2020) 101–109.

  • Tarvainen A & Valpola H, Mean teachers are better role models: Weight-averaged consistency targets improve semisupervised deep learning results, arXiv Prepr arXiv170301780, (2017).

  • Miyato T, Maeda S, Koyama M & Ishii S, Virtual adversarial training: a regularization method for supervised and semisupervised learning, IEEE Trans Pattern Anal Mach Intell, 41(8) (2018) 1979–1993, https://doi.org/10.1109/ TPAMI.2018.2858821.

  • Berthelot D, Carlini N, Goodfellow I, Papernot N, Oliver A & Raffel C, Mixmatch: A holistic approach to semisupervised learning, arXiv Prepr arXiv190502249, Published online (2019).

  • Howard A G, Zhu M, Chen B, Kalenichenko D, Wang W, Weyand T, Andreetto M & Adam H, Mobilenets: Efficient convolutional neural networks for mobile vision applications, arXiv Prepr arXiv170404861, (2017).

  • He K, Zhang X, Ren S & Sun J, Identity mappings in deep residual networks, in Computer Vision–ECCV 2016: 14th Euro Conf, Amsterdam, The Netherlands, October 11–14, 2016, Proc, Part IV 14 (Springer International Publishing 2016, 630–645, http://arxiv.org/abs/1603.05027.

  • He K, Zhang X, Ren S & Sun J, Deep residual learning for image recognition, CoRR, (2015) abs/1512.0, http://arxiv.org/abs/1512.03385.

  • Zhang X, Zou J, He K & Sun J, Accelerating very deep convolutional networks for classification and detection, IEEE Trans Pattern Anal Mach Intell, 38(10) (2015) 1943–1955.

  • Simonyan K & Zisserman A, Very deep convolutional networks for large-scale image recognition, arXiv Prepr arXiv14091556,(2014).

  • Setiawan F, Yahya B N & Lee S L, Deep activity recognition on imaging sensor data, Electron Lett, 55(17) (2019) 28–931.

  • Benjamini Y, Discovering the false discovery rate, J R Stat Soc Ser B statistical Methodol, 72(4) (2010) 405–416.

  • Mall P K, Singh P K & Yadav D, GLCM based feature extraction and medical X-RAY image classification using machine learning techniques, in 2019 IEEE Conf Info Commun Technol(IEEE) 2019, 1–6.

  • Narayan V & Daniel A K, CHOP: Maximum coverage optimization and resolve hole healing problem using sleep and wake-up technique for WSN, Adv Distrib Comput Artif Intell J, 11(2) (2022) 159–178.

  • Narayan V & Daniel A K, RBCHS: Region-based cluster head selection protocol in wireless sensor network, in Proc Integrat Intell Enable Netw Comput (Springer) 2021, 863–869.

  • Srivastava S & Sharma S, Analysis of cyber related issues by implementing data mining algorithm, in 2019 9th Int Conf Cloud Comput Data Sci Eng (IEEE) 2019, 606–610, https://doi.org/10.1109/CONFLUENCE.2019.8776980.

  • Chicco D & Jurman G, The advantages of the Matthews correlation coefficient (MCC) over F1 score and accuracy in binary classification evaluation, BMC Genomics, 21(1) (2020) 1–13.


Abstract Views: 23

PDF Views: 18




  • Rank Based Two Stage Semi-Supervised Deep Learning Model for X-Ray Images Classification

Abstract Views: 23  |  PDF Views: 18

Authors

Pawan Kumar Mall
Madan Mohan Malaviya University of Technology, Gorakhpur 273 016, Uttar Pradesh, India
Vipul Narayan
Galgotias University, Greater Noida 203 201, Uttar Pradesh, India
Swapnita Srivastava
Galgotias University, Greater Noida 203 201, Uttar Pradesh, India
Munish Sabarwal
Galgotias University, Greater Noida 203 201, Uttar Pradesh, India
Vimal Kumar
Galgotias University, Greater Noida 203 201, Uttar Pradesh, India
Shashank Awasthi
GL Bajaj Institute of Technology and Management, Greater Noida203 201, Uttar Pradesh, India
Lalit Tyagi
GL Bajaj Institute of Technology and Management, Greater Noida203 201, Uttar Pradesh, India

Abstract


Deep learning approaches rely on a wide-scale labeled dataset to attain a high level of performance. Although labeled data is more difficult and costly to access in some applications, such as bioinformatics and medical imaging, wide variety of ongoing research on the topic of Semi-Supervised Deep Learning (SSDL) can improve and fix underlying problems in this domain. The motivation for the suggested model Rank Based Two-Stage Semi-Supervised Deep Learning (RTS-SS-DL) is the same as how doctors deal with unobserved or suspect cases in day to day practice. The physicians deal with these suspect instances with the help of professional assistance from their colleagues. Before beginning therapy, some patients seek the opinion of a variety of skilled professionals. The patients are treated by the most appropriate (vote count) professional diagnosis. Our model (RTS-SS-DL) has achieved impressive metrics including 92.776% accuracy, 97.376% specificity, 86.932% sensitivity, 96.192% precision, 85.644% MCC (Matthews Correlation Coefficient), 3.808% FDR (False Discovery Rate), 2.624% FPR (False Positive Rate), 91.072% f1-score, 90.85% NPV (Negative Predictive Value), and 13.068% FNR (False Negative Rate) for the unseen dataset. The outcome of this research results in an SSDL model that is both more precise and effective.

Keywords


Labeled dataset, RTS-SS-DL, Self-organising classifier, Semi-supervised learning, Shoulder’s fracture classification.

References