Explainable Machine Learning Techniques in Medical Image Analysis Based on Classification with Feature Extraction

Aug 4, 2026

Upload by Jiven

The document presents an explainable machine learning approach for COVID-19 lung infection detection using medical image analysis. A COVID patient lung image dataset is collected and processed through noise removal and image smoothening, then spatio transfer neural network features are extracted with a DenseNet+ integrated pipeline. The extracted representations are classified via a stacked auto Boltzmann encoder combined with VGG-19Net+ under transfer learning for binary classification. Experiments across COVID datasets evaluate accuracy, precision, recall, F1-score, RMSE, and MAP, achieving 95% accuracy, 91% precision, 85% recall, 80% F1-score, 61% RMSE, and 51% MAP.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages16

Reading Time40 min

AccessFree to download

Introduction

Research background and motivation

Contribution of the research

Explainable Machine Learning Techniques in Medical Image Analysis Based on Classification with Feature Extraction Dr. B. Dwarakanath Assistant Professor, Department of Information Technology, SRM Institute of Science and Technology, Ramapuram, Chennai, India [dwarakab@srmist.edu.in](dwarakab@srmist.edu.in) [Dr](Dr). Gitanjali Shrivastava Assistant Professor, Department of Law, Symbiosis Law School, Pune, India [dr.gitanjali10@gmail.com](dr.gitanjali10@gmail.com) [Dr](Dr). Rohit Bansal Associate Professor, Department of Management Studies, Vaish College of Engineering, Rohtak, Haryana, India [rohitbansal.mba@gmail.com](rohitbansal.mba@gmail.com) Praful Nandankar Assistant Professor, Department of Electrical Engineering, Government College of Engineering, Nagpur, India [pppful@gmail.com](pppful@gmail.com) [Dr](Dr). Veera Talukdar Registrar, Kaziranga University, Jorhat, Assam, India [bhaskarveera95@gmail.com](bhaskarveera95@gmail.com) M Ahmer Usmani Assistant Professor, Department of Computer Science and Engineering, Bharati Vidyapeeth Deemed to be University, Department of Engineering and Technology, Navi Mumbai, India [mausmani@bvucoep.edu.in](mausmani@bvucoep.edu.in) | Article History
Received: 13 July 2022
Revised: 20 September 2022
Accepted: 26 October 2022 | Abstract
Animals are also afflicted by COVID-19, a virus that is quickly spreading and infects both humans and animals. This fatal viral disease has an impact on people's daily lives, health, and economy of a nation. Most effective machine learning method is deep learning, which offers insightful analysis for examining a significant number of chest x-ray pictures that have a significant bearing on COVID-19 screening. This research proposes novel technique in lung image analysis for detection of lung infection due to COVID using Explainable Machine learning techniques. Here the input has been collected as COVID patient’s lung image dataset and it has been processed for noise removal and smoothening. This processed image features have been extracted using spatio transfer neural network integrated with DenseNet+ architecture. Extracted features has been classified using stacked auto Boltzmann encoder machine with VGG- 19Net+ . With the transfer learning method integrated into the binary classification process, the suggested algorithm achieves good classification accuracy. The experimental analysis has been carried out for various COVID dataset in terms of accuracy, precision, Recall, F-1score, RMSE, MAP. The proposed technique attained accuracy of 95%, precision of 91%, recall of 85%, F_ 1 score of 80%, RMSE of 61% and MAP of 51% . | | --- | --- | | CC License
CC-BY-NC-SA 4. | Keywords: lung image analysis, lung infection, Explainable Machine learning, classification, COVID-19, feature extraction. | | --- | --- | 1. Introduction Covid-19 is a serious medical problem where many individuals pass away every day. This virus sickness not only impacts a particular nation, but also causes suffering throughout the entire planet. Several different viruses (including SARS, MERS, the flu, etc.) [1] entered the scene in the last ten years, although they only last a few days or weeks. Due to the availability of vaccinations created by these experts, few of these viruses are diagnosed even though many scientists are working on them. The Covid-19 sickness is currently affecting everyone in the world [2], and the most crucial issue is that no one country's experts have been able to develop a vaccine for it. Plasma therapy, X-ray imaging, and many more forecasts came into play in the meantime, but no precise cure for this fatal illness has been discovered. Covid-19 [3] is a disease that claims lives of individuals every day, and it is exceedingly expensive to diagnose for a nation, a state, and a patient. Online repositories like Githuband Kaggle had X-ray photos of healthy persons and Covid-19-infected patients [4] for stud

FAQs

What problem does the proposed method address?

It targets detecting lung infection due to COVID-19 from medical images, using explainable machine learning techniques for classification.

How are input lung images prepared and features extracted?

The lung images are collected from COVID patient datasets, processed with noise removal and smoothening, and then features are extracted using a spatio transfer neural network integrated with DenseNet+.

Which model components are used for classification, and what performance results are reported?

Features are classified using a stacked auto Boltzmann encoder with VGG-19Net+ under transfer learning for binary classification. Reported results include 95% accuracy, 91% precision, 85% recall, 80% F1-score, 61% RMSE, and 51% MAP.

Aug 4, 2026

Upload by Jiven