Sathiya Kumar Chinnappan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sathiya Kumar Chinnappan has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
Syed Mohammed Shafi
School of Computer Science and Engineering, Vellore Institute of Technology, Vellore, India
3 shared publications- Sudhakar Sengan
Department of Computer Science and Engineering, Erode Sengunthar Engineering College, Erode, India
2 shared publications - Anil Kumar Dubey
ICAR-Central Soil Salinity Research Institute, Regional Research Station, Lucknow, India
1 shared publication - Arjun Subburaj
Tranxit Technology Solutions Private Limited
1 shared publication - C. R. Dhivyaa
School of Computer Science and Engineering, Vellore Institute of Technology, Vellore, Tamil Nadu, India
1 shared publication - Dheeba Jacob
Vellore Institute of Technology
1 shared publication
Latest Publications
- 1 March 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (5) , 055226
Multimodal diagnosis: co-attention transformer for accurate pneumonia classification from chest radiographs and clinical reports
Chest x-rays are tricky; the fine details and overlaps make it difficult for doctors to diagnose lung issues quickly and accurately. Image-only reading has limitations because features look too similar, and not everyone spots them the same way. Old methods also fall short at blending pictures with reports, so the full story between visuals and words gets lost. The CoATr architecture employs co-attention mechanisms and transformer models to improve cross-modal interactions while boosting multimodal task performance. The system builds a framework that unifies visual chest radiographic data with clinical report semantic information to achieve accurate multi-label classification of chest diseases. The model uses a Swin Transformer backbone to extract detailed visual information from thoracic radiographs. ClinicalBERT was used for clinical report analysis to create contextualised text embeddings. The co-attention mechanism functions to create a shared model which enables visual and textual elements to interact, thus improving their performance through mutual enhancement. The CoATr model was evaluated using the NIH ChestX-ray14 dataset, which contains over 1,12,000 frontal chest radiographs annotated with 14 thoracic disease labels and the associated clinical text. In our tests, the model performed better than standard CNNs and single-mode transformers, and the AUC, precision, and recall all improved noticeably. Since it handles images and text together while explaining its decisions, hospitals can use it to auto-classify chest diseases.
- 1 March 2026
- Published by IOP Publishing in Engineering Research Express
- Vol. 8 (5) , 055226
Multimodal diagnosis: co-attention transformer for accurate pneumonia classification from chest radiographs and clinical reports
Chest x-rays are tricky; the fine details and overlaps make it difficult for doctors to diagnose lung issues quickly and accurately. Image-only reading has limitations because features look too similar, and not everyone spots them the same way. Old methods also fall short at blending pictures with reports, so the full story between visuals and words gets lost. The CoATr architecture employs co-attention mechanisms and transformer models to improve cross-modal interactions while boosting multimodal task performance. The system builds a framework that unifies visual chest radiographic data with clinical report semantic information to achieve accurate multi-label classification of chest diseases. The model uses a Swin Transformer backbone to extract detailed visual information from thoracic radiographs. ClinicalBERT was used for clinical report analysis to create contextualised text embeddings. The co-attention mechanism functions to create a shared model which enables visual and textual elements to interact, thus improving their performance through mutual enhancement. The CoATr model was evaluated using the NIH ChestX-ray14 dataset, which contains over 1,12,000 frontal chest radiographs annotated with 14 thoracic disease labels and the associated clinical text. In our tests, the model performed better than standard CNNs and single-mode transformers, and the AUC, precision, and recall all improved noticeably. Since it handles images and text together while explaining its decisions, hospitals can use it to auto-classify chest diseases.
- Open Access
- 13 December 2024
- Published by PeerJ in PeerJ Computer Science
- Vol. 10, e2444
Hybrid transformer-CNN and LSTM model for lung disease segmentation and classification
According to the World Health Organization (WHO) report, lung disorders are the third leading cause of mortality worldwide. Approximately three million individuals are affected with various types of lung disorders annually. This issue alarms us to take control measures related to early diagnostics, accurate treatment procedures, etc. The precise identification through the assessment of medical images is crucial for pulmonary disease diagnosis. Also, it remains a formidable challenge due to the diverse and unpredictable nature of pathological lung appearances and shapes. Therefore, the efficient lung disease segmentation and classification model is essential. By taking this initiative, a novel lung disease segmentation with a hybrid LinkNet-Modified LSTM (L-MLSTM) model is proposed in this research article. The proposed model utilizes four essential and fundamental steps for its implementation. The first step is pre-processing, where the input lung images are pre-processed using median filtering. Consequently, an improved Transformer-based convolutional neural network (CNN) model (ITCNN) is proposed to segment the affected region in the segmentation process. After segmentation, essential features such as texture, shape, color, and deep features are retrieved. Specifically, texture features are extracted using modified Local Gradient Increasing Pattern (LGIP) and Multi-texton analysis. Then, the classification step utilizes a hybrid model, the L-MLSTM model. This work leverages two datasets such as the COVID-19 normal pneumonia-CT images dataset (Dataset 1) and the Chest CT scan images dataset (Dataset 2). The dataset is crucial for training and evaluating the model, providing a comprehensive basis for robust and generalizable results. The L-MLSTM model outperforms several existing models, including HDE-NN, DBN, LSTM, LINKNET, SVM, Bi-GRU, RNN, CNN, and VGG19 + CNN, with accuracies of 89% and 95% at learning percentages of 70 and 90, respectively, for datasets 1 and 2. The improved accuracy achieved by the L-MLSTM model highlights its capability to better handle the complexity and variability in lung images. This hybrid approach enhances the model’s ability to distinguish between different types of lung diseases and reduces diagnostic errors compared to existing methods.
- 13 December 2024
- Published by PeerJ in PeerJ Computer Science
- Vol. 10, e2444
Open AccessHybrid transformer-CNN and LSTM model for lung disease segmentation and classification
According to the World Health Organization (WHO) report, lung disorders are the third leading cause of mortality worldwide. Approximately three million individuals are affected with various types of lung disorders annually. This issue alarms us to take control measures related to early diagnostics, accurate treatment procedures, etc. The precise identification through the assessment of medical images is crucial for pulmonary disease diagnosis. Also, it remains a formidable challenge due to the diverse and unpredictable nature of pathological lung appearances and shapes. Therefore, the efficient lung disease segmentation and classification model is essential. By taking this initiative, a novel lung disease segmentation with a hybrid LinkNet-Modified LSTM (L-MLSTM) model is proposed in this research article. The proposed model utilizes four essential and fundamental steps for its implementation. The first step is pre-processing, where the input lung images are pre-processed using median filtering. Consequently, an improved Transformer-based convolutional neural network (CNN) model (ITCNN) is proposed to segment the affected region in the segmentation process. After segmentation, essential features such as texture, shape, color, and deep features are retrieved. Specifically, texture features are extracted using modified Local Gradient Increasing Pattern (LGIP) and Multi-texton analysis. Then, the classification step utilizes a hybrid model, the L-MLSTM model. This work leverages two datasets such as the COVID-19 normal pneumonia-CT images dataset (Dataset 1) and the Chest CT scan images dataset (Dataset 2). The dataset is crucial for training and evaluating the model, providing a comprehensive basis for robust and generalizable results. The L-MLSTM model outperforms several existing models, including HDE-NN, DBN, LSTM, LINKNET, SVM, Bi-GRU, RNN, CNN, and VGG19 + CNN, with accuracies of 89% and 95% at learning percentages of 70 and 90, respectively, for datasets 1 and 2. The improved accuracy achieved by the L-MLSTM model highlights its capability to better handle the complexity and variability in lung images. This hybrid approach enhances the model’s ability to distinguish between different types of lung diseases and reduces diagnostic errors compared to existing methods.
- Open Access
- 25 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 176920-176937
Explainable Model of Fusion Network With Enhanced Optimization Approach for Tuberculosis Diagnosis
Tuberculosis, caused by Mycobacterium tuberculosis, is a lung-focused infectious disease with a profound impact on the respiratory system. Although advancements in healthcare have improved detection of various diseases, TB remains a global health issue, especially in developing regions. Efficient TB diagnostics can be achieved through the deep-learning-based image classification techniques. In this work, a novel hybrid model is introduced in which two approaches are proposed. In the first approach, it integrates three Convolutional models known as ResNet-50, InceptionV3, and EfficientB4 to detect deep features. After deep feature set creation, Principal Component Analysis is employed to decrease the feature dimensionality. The second approach utilizes the EAMSO (Ensemble of AMS Optimization model) for feature selection. It combines the results from AEO (Artificial Ecosystem-based Optimization), MBO (Monarch butterfly optimization), and Seagull-Algorithm to form three optimized feature subsets. These subsets are correlated by ensemble model, and the optimal features from EAMSO are input into an Artificial Neural Network for classification. The various models are compared for analyzing the performance of the classification using two benchmark CXR image datasets. The proposed model yields a remarkable average accuracy value of 98.80%. Furthermore, explainable Artificial Intelligence experiments validate the effectiveness of the proposed diagnostic approach.
- 25 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 176920-176937
Open AccessExplainable Model of Fusion Network With Enhanced Optimization Approach for Tuberculosis Diagnosis
Tuberculosis, caused by Mycobacterium tuberculosis, is a lung-focused infectious disease with a profound impact on the respiratory system. Although advancements in healthcare have improved detection of various diseases, TB remains a global health issue, especially in developing regions. Efficient TB diagnostics can be achieved through the deep-learning-based image classification techniques. In this work, a novel hybrid model is introduced in which two approaches are proposed. In the first approach, it integrates three Convolutional models known as ResNet-50, InceptionV3, and EfficientB4 to detect deep features. After deep feature set creation, Principal Component Analysis is employed to decrease the feature dimensionality. The second approach utilizes the EAMSO (Ensemble of AMS Optimization model) for feature selection. It combines the results from AEO (Artificial Ecosystem-based Optimization), MBO (Monarch butterfly optimization), and Seagull-Algorithm to form three optimized feature subsets. These subsets are correlated by ensemble model, and the optimal features from EAMSO are input into an Artificial Neural Network for classification. The various models are compared for analyzing the performance of the classification using two benchmark CXR image datasets. The proposed model yields a remarkable average accuracy value of 98.80%. Furthermore, explainable Artificial Intelligence experiments validate the effectiveness of the proposed diagnostic approach.
- Open Access
- 16 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0302507
Segmenting and classifying lung diseases with M-Segnet and Hybrid Squeezenet-CNN architecture on CT images
Diagnosing lung diseases accurately and promptly is essential for effectively managing this significant public health challenge on a global scale. This paper introduces a new framework called Modified Segnet-based Lung Disease Segmentation and Severity Classification (MSLDSSC). The MSLDSSC model comprises four phases: "preprocessing, segmentation, feature extraction, and classification." Initially, the input image undergoes preprocessing using an improved Wiener filter technique. This technique estimates the power spectral density of the noisy and original images and computes the SNR assisted by PSNR to evaluate image quality. Next, the preprocessed image undergoes Segmentation to identify and separate the RoI from the background objects in the lung image. We employ a Modified Segnet mechanism that utilizes a proposed hard tanh-Softplus activation function for effective Segmentation. Following Segmentation, features such as MLDN, entropy with MRELBP, shape features, and deep features are extracted. Following the feature extraction phase, the retrieved feature set is input into a hybrid severity classification model. This hybrid model comprises two classifiers: SDPA-Squeezenet and DCNN. These classifiers train on the retrieved feature set and effectively classify the severity level of lung diseases.
- 16 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0302507
Open AccessSegmenting and classifying lung diseases with M-Segnet and Hybrid Squeezenet-CNN architecture on CT images
Diagnosing lung diseases accurately and promptly is essential for effectively managing this significant public health challenge on a global scale. This paper introduces a new framework called Modified Segnet-based Lung Disease Segmentation and Severity Classification (MSLDSSC). The MSLDSSC model comprises four phases: "preprocessing, segmentation, feature extraction, and classification." Initially, the input image undergoes preprocessing using an improved Wiener filter technique. This technique estimates the power spectral density of the noisy and original images and computes the SNR assisted by PSNR to evaluate image quality. Next, the preprocessed image undergoes Segmentation to identify and separate the RoI from the background objects in the lung image. We employ a Modified Segnet mechanism that utilizes a proposed hard tanh-Softplus activation function for effective Segmentation. Following Segmentation, features such as MLDN, entropy with MRELBP, shape features, and deep features are extracted. Following the feature extraction phase, the retrieved feature set is input into a hybrid severity classification model. This hybrid model comprises two classifiers: SDPA-Squeezenet and DCNN. These classifiers train on the retrieved feature set and effectively classify the severity level of lung diseases.
- Open Access
- 11 August 2022
- Published by Wiley in Mathematical Problems in Engineering
- Vol. 2022, 1-7
Fuzzy Logic and Machine Learning-Enabled Recommendation System to Predict Suitable Academic Program for Students
In recent years, educational data mining has gained a considerable lot of interest as a consequence of the large number of pedagogical content that can be gathered from a range of sources. This is because there is a lot of instructional information that can be obtained. The data mining tools collaborate with academics to improve students’ learning strategies by analyzing, sifting through, and estimating components that are pertinent to students’ characteristics or patterns of behavior. This is accomplished through the following steps: EDM is utilized in the vast majority of instances to develop the classification model, which then assigns a certain class to each student based on the known properties of the training dataset. Before putting the classification model into use, it is possible to utilize a test dataset to verify that the model is accurate. This article provides a description of a recommendation system that determines the most beneficial academic program for students by utilizing fuzzy logic and machine learning. The compilation of a student dataset has begun. It includes a total of 21 features and 1000 individual cases. The initial step is to employ the CFS attribute selection method. This methodology selects 15 of the initial set of 21 characteristics. Following the completion of the data gathering, it is put through various machine learning methods such fuzzy SVM, random forest, and C4.5. This methodology that has been offered makes predictions about the academic program that is best suitable for students.
- 11 August 2022
- Published by Wiley in Mathematical Problems in Engineering
- Vol. 2022, 1-7
Open AccessFuzzy Logic and Machine Learning-Enabled Recommendation System to Predict Suitable Academic Program for Students
In recent years, educational data mining has gained a considerable lot of interest as a consequence of the large number of pedagogical content that can be gathered from a range of sources. This is because there is a lot of instructional information that can be obtained. The data mining tools collaborate with academics to improve students’ learning strategies by analyzing, sifting through, and estimating components that are pertinent to students’ characteristics or patterns of behavior. This is accomplished through the following steps: EDM is utilized in the vast majority of instances to develop the classification model, which then assigns a certain class to each student based on the known properties of the training dataset. Before putting the classification model into use, it is possible to utilize a test dataset to verify that the model is accurate. This article provides a description of a recommendation system that determines the most beneficial academic program for students by utilizing fuzzy logic and machine learning. The compilation of a student dataset has begun. It includes a total of 21 features and 1000 individual cases. The initial step is to employ the CFS attribute selection method. This methodology selects 15 of the initial set of 21 characteristics. Following the completion of the data gathering, it is put through various machine learning methods such fuzzy SVM, random forest, and C4.5. This methodology that has been offered makes predictions about the academic program that is best suitable for students.
- 20 May 2022
- Published by Springer Nature in Wireless Personal Communications
- Vol. 125 (4) , 3205-3225
A Smart Parking System: An IoT Based Computer Vision Approach for Free Parking Spot Detection Using Faster R-CNN with YOLOv3 Method
- 20 May 2022
- Published by Springer Nature in Wireless Personal Communications
- Vol. 125 (4) , 3205-3225
A Smart Parking System: An IoT Based Computer Vision Approach for Free Parking Spot Detection Using Faster R-CNN with YOLOv3 Method
- Retracted
- 13 April 2021
- Published by Springer Nature in Arabian Journal for Science and Engineering
- Vol. 48 (2) , 2589
RETRACTED ARTICLE: Prewitt Logistic Deep Recurrent Neural Learning for Face Log Detection by Extracting Features from Images
- 13 April 2021
- Published by Springer Nature in Arabian Journal for Science and Engineering
- Vol. 48 (2) , 2589
RetractedRETRACTED ARTICLE: Prewitt Logistic Deep Recurrent Neural Learning for Face Log Detection by Extracting Features from Images