Hafiz Muhammad Tayyab Khushi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hafiz Muhammad Tayyab Khushi has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Tehreem Masood
Faculty of Computer Science and Information Technology, Superior University, Lahore, Pakistan
5 shared publications Arfan Jaffar
Superior University
3 shared publications- H. M. Shahzad
Faculty of Computer Science and Information Technology, The Superior University, Lahore 54000, Pakistan
3 shared publications - Hassan A. Ahmed
Information Systems Department, Cleveland State University, Cleveland, OH 44114, USA
3 shared publications - Sheeraz Akram
The Superior University
3 shared publications Syed Hamza Ahmed
Cleveland State University
2 shared publications
Latest Publications
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 147842
Corrections to “HistoDX: Revolutionizing Breast Cancer Diagnosis Through Advanced Imaging Techniques”
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 14, 147842
Corrections to “HistoDX: Revolutionizing Breast Cancer Diagnosis Through Advanced Imaging Techniques”
- Open Access
- 25 June 2025
- Published by Springer Nature in Research Square
Optimizer-Aware Deep Learning for Brain Tumor Classification: A Study Using AlexNet to EfficientNetB0
One of the most lethal diseases in the world is brain tumors. Convolutional neural networks (CNNs) and other deep learning-based methods are vital for accurate diagnosis because of their remarkable capacity for learning and prediction. These techniques are widely used in image processing, computer vision, and medical diagnostic tasks such as classification, segmentation, and object detection. In this study, we use the publicly available Figshare brain tumor dataset to compare the performance of different pre-trained deep learning models, including AlexNet, VGG16, VGG19, ResNet50, Xception, InceptionV3, DenseNet121, MobileNetV1, and EfficientNetB0, on four optimizers, such as adaptive moment estimation (Adam), stochastic gradient descent (SGD), root mean square propagation (RMSProp), and Nesterov-accelerated adaptive moment estimation (Nadam). The experimental results show that the VGG19 and EfficientNetB0 models performed exceptionally well with the Adam optimizer, achieving an overall accuracy of 99.13% and a misclassification rate of 0.87%. Additionally, receiver operating characteristic (ROC) curves were calculated, with the EfficientNetB0 model achieving an area under the curve (AUC) value of 100% for each class. It also demonstrated excellent performance on the test images, with a testing accuracy of 99.61%.
- 25 June 2025
- Published by Springer Nature in Research Square
Open AccessOptimizer-Aware Deep Learning for Brain Tumor Classification: A Study Using AlexNet to EfficientNetB0
One of the most lethal diseases in the world is brain tumors. Convolutional neural networks (CNNs) and other deep learning-based methods are vital for accurate diagnosis because of their remarkable capacity for learning and prediction. These techniques are widely used in image processing, computer vision, and medical diagnostic tasks such as classification, segmentation, and object detection. In this study, we use the publicly available Figshare brain tumor dataset to compare the performance of different pre-trained deep learning models, including AlexNet, VGG16, VGG19, ResNet50, Xception, InceptionV3, DenseNet121, MobileNetV1, and EfficientNetB0, on four optimizers, such as adaptive moment estimation (Adam), stochastic gradient descent (SGD), root mean square propagation (RMSProp), and Nesterov-accelerated adaptive moment estimation (Nadam). The experimental results show that the VGG19 and EfficientNetB0 models performed exceptionally well with the Adam optimizer, achieving an overall accuracy of 99.13% and a misclassification rate of 0.87%. Additionally, receiver operating characteristic (ROC) curves were calculated, with the EfficientNetB0 model achieving an area under the curve (AUC) value of 100% for each class. It also demonstrated excellent performance on the test images, with a testing accuracy of 99.61%.
- Open Access
- 27 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 94416-94436
HistoDX: Revolutionizing Breast Cancer Diagnosis Through Advanced Imaging Techniques
Breast cancer is the second leading cause of mortality among women worldwide, highlighting the need for efficient histopathology-based screening methods for early diagnosis. This study introduces HistoDX, a deep learning framework to classify Invasive Ductal Carcinoma (IDC) using 277,524 histopathology image patches (50x50 pixels) from Paul Mooney’s IDC dataset on Kaggle, comprising No Cancer and IDC(+) classes. HistoDX employs a preprocessing pipeline with normalization, data augmentation, and class balancing via oversampling and weighted loss to address the class imbalance. A customized convolutional neural network, built on EfficientNetV2-B3 with additional layers, achieves 97% accuracy and a 0.91 ROC-AUC score on the test set. Validation on BreakHis (97% accuracy) and BACH (90% accuracy) datasets confirm generalizability, though detecting minority IDC(+) cases remains challenging. Low training and test losses underscore reliability. HistoDX empowers pathologists by enhancing diagnostic efficiency and minimizing subjectivity through effective class imbalance mitigation. Its robust performance across diverse datasets like BreakHis and BACH suggests readiness for clinical integration. Future research into advanced augmentation techniques, ensemble models, and whole-slide image analysis could further optimize accuracy, sensitivity, and scalability, paving the way for broader adoption in precision oncology.
- 27 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 94416-94436
Open AccessHistoDX: Revolutionizing Breast Cancer Diagnosis Through Advanced Imaging Techniques
Breast cancer is the second leading cause of mortality among women worldwide, highlighting the need for efficient histopathology-based screening methods for early diagnosis. This study introduces HistoDX, a deep learning framework to classify Invasive Ductal Carcinoma (IDC) using 277,524 histopathology image patches (50x50 pixels) from Paul Mooney’s IDC dataset on Kaggle, comprising No Cancer and IDC(+) classes. HistoDX employs a preprocessing pipeline with normalization, data augmentation, and class balancing via oversampling and weighted loss to address the class imbalance. A customized convolutional neural network, built on EfficientNetV2-B3 with additional layers, achieves 97% accuracy and a 0.91 ROC-AUC score on the test set. Validation on BreakHis (97% accuracy) and BACH (90% accuracy) datasets confirm generalizability, though detecting minority IDC(+) cases remains challenging. Low training and test losses underscore reliability. HistoDX empowers pathologists by enhancing diagnostic efficiency and minimizing subjectivity through effective class imbalance mitigation. Its robust performance across diverse datasets like BreakHis and BACH suggests readiness for clinical integration. Future research into advanced augmentation techniques, ensemble models, and whole-slide image analysis could further optimize accuracy, sensitivity, and scalability, paving the way for broader adoption in precision oncology.
- 29 March 2025
- Published by Springer Nature
- p. 149-170
Deep-Learning Based Multi-Modalities Fusion for the Detection of Brain-Related Diseases: A Review
- 29 March 2025
- Published by Springer Nature
- p. 149-170
Deep-Learning Based Multi-Modalities Fusion for the Detection of Brain-Related Diseases: A Review
- Open Access
Benchmarking Pretrained Models for Speech Emotion Recognition: A Focus on Xception
Speech emotion recognition (SER) is an emerging technology that utilizes speech sounds to identify a speaker’s emotional state. Computational intelligence is receiving increasing attention from academics, health, and social media applications. This research was conducted to identify emotional states in verbal communication. We applied a publicly available dataset called RAVDEES. The data augmentation process involved adding noise, applying time stretching, shifting, and pitch, and extracting the features zero cross rate (ZCR), chroma shift, Mel-Frequency Cepstral Coefficients (MFCC), and a spectrogram. In addition, we used many pretrained deep learning models, such as VGG16, ResNet50, Xception, InceptionV3, and DenseNet121. Out of all of the deep learning models, Xception yielded superior outcomes. Furthermore, we improved performance by changing the Xception model to include hyperparameters and additional layers. We used a variety of performance evaluation parameters to test the proposed model. These included F1-score, accuracy, misclassification rate (MCR), precision, sensitivity, specificity, negative predictive value, false negative rate, false positive rate, false discovery rate, false omission rate, and false discovery rate. The model that we suggested demonstrated an overall accuracy of 98%, with an MCR of 2%. Additionally, it attained precision, sensitivity, and specificity values of 91.99%, 91.78%, and 98.68%, respectively. Additional models attained an F1-score of 91.83%. Our suggested model demonstrated superiority compared to other cutting-edge techniques
Open AccessBenchmarking Pretrained Models for Speech Emotion Recognition: A Focus on Xception
Speech emotion recognition (SER) is an emerging technology that utilizes speech sounds to identify a speaker’s emotional state. Computational intelligence is receiving increasing attention from academics, health, and social media applications. This research was conducted to identify emotional states in verbal communication. We applied a publicly available dataset called RAVDEES. The data augmentation process involved adding noise, applying time stretching, shifting, and pitch, and extracting the features zero cross rate (ZCR), chroma shift, Mel-Frequency Cepstral Coefficients (MFCC), and a spectrogram. In addition, we used many pretrained deep learning models, such as VGG16, ResNet50, Xception, InceptionV3, and DenseNet121. Out of all of the deep learning models, Xception yielded superior outcomes. Furthermore, we improved performance by changing the Xception model to include hyperparameters and additional layers. We used a variety of performance evaluation parameters to test the proposed model. These included F1-score, accuracy, misclassification rate (MCR), precision, sensitivity, specificity, negative predictive value, false negative rate, false positive rate, false discovery rate, false omission rate, and false discovery rate. The model that we suggested demonstrated an overall accuracy of 98%, with an MCR of 2%. Additionally, it attained precision, sensitivity, and specificity values of 91.99%, 91.78%, and 98.68%, respectively. Additional models attained an F1-score of 91.83%. Our suggested model demonstrated superiority compared to other cutting-edge techniques
- Open Access
- 1 January 2024
- Published by FapUNIFESP (SciELO) in Brazilian Archives of Biology and Technology
- Vol. 67, e24231137
A Novel Approach to Classify Brain Tumor with an Effective Transfer Learning based Deep Learning Model
World's deadliest disease is brain tumor. Misdiagnosed cancers and inadequate treatment reduce survival. However, magnetic resonance imaging (MRI) is utilized for tumor analysis, but the enormous number of pictures produced by MRI makes it time-consuming and difficult to diagnose a patient just because of its complex nature, putting their life at risk. Thus, accurately detecting early brain cancers manually is difficult. We need an autonomous, intelligent system to detect brain cancers early and accurately. This study proposes a pre-trained EfficientNetb4 model with an adjusttable learning rate and custom callback to efficiently classify tumors. The proposed methodology improves the quality and quantity of the publicly accessible Br35h dataset by applying data augmentation techniques. The suggested model had 99.67% accuracy on Br35h data and 0.33% miss categorization. This method, however, had 99.87% accuracy on supplemented data and 0.13% miss categorization. The system achieves 99.33%, 99.97%, 99.93%, 99.33%, 0.66%, and 99.66% for Br35h in sensitivity, specificity, precision, NPV, FOR, and F1-score. The suggested model had 99.97% sensitivity, 99.74% specificity, 99.74% accuracy, 100% NPV, 0.66% FOR, and 99.66% F1-score for the Br35h-augmented dataset. Further, the proposed model also achieved FNR and FPR of 0 and 0.26%, respectively, and the augmented Br35H dataset achieved FNR and FPR of 0.66% and 0%, respectively. The 5Flod cross-validation also found that the model achieved an overall validation accuracy of 98.606% with a loss of 0.224%. Regarding results, the proposed system shows superior performance to other state-of-the art approaches.
- 1 January 2024
- Published by FapUNIFESP (SciELO) in Brazilian Archives of Biology and Technology
- Vol. 67, e24231137
Open AccessA Novel Approach to Classify Brain Tumor with an Effective Transfer Learning based Deep Learning Model
World's deadliest disease is brain tumor. Misdiagnosed cancers and inadequate treatment reduce survival. However, magnetic resonance imaging (MRI) is utilized for tumor analysis, but the enormous number of pictures produced by MRI makes it time-consuming and difficult to diagnose a patient just because of its complex nature, putting their life at risk. Thus, accurately detecting early brain cancers manually is difficult. We need an autonomous, intelligent system to detect brain cancers early and accurately. This study proposes a pre-trained EfficientNetb4 model with an adjusttable learning rate and custom callback to efficiently classify tumors. The proposed methodology improves the quality and quantity of the publicly accessible Br35h dataset by applying data augmentation techniques. The suggested model had 99.67% accuracy on Br35h data and 0.33% miss categorization. This method, however, had 99.87% accuracy on supplemented data and 0.13% miss categorization. The system achieves 99.33%, 99.97%, 99.93%, 99.33%, 0.66%, and 99.66% for Br35h in sensitivity, specificity, precision, NPV, FOR, and F1-score. The suggested model had 99.97% sensitivity, 99.74% specificity, 99.74% accuracy, 100% NPV, 0.66% FOR, and 99.66% F1-score for the Br35h-augmented dataset. Further, the proposed model also achieved FNR and FPR of 0 and 0.26%, respectively, and the augmented Br35H dataset achieved FNR and FPR of 0.66% and 0%, respectively. The 5Flod cross-validation also found that the model achieved an overall validation accuracy of 98.606% with a loss of 0.224%. Regarding results, the proposed system shows superior performance to other state-of-the art approaches.
- Open Access
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 117210-117230
Improved Multiclass Brain Tumor Detection via Customized Pretrained EfficientNetB7 Model
A brain tumor considered the deadliest disease in the world. Patients with misdiagnoses and insufficient treatment have a lower chance of surviving for life. However, for diagnosing the disturbance in the brain, magnetic resonance images play a vital role, but due to the large number of images produced by MRI, it is time-consuming and difficult to diagnose a patient at an earlier stage with high accuracy. Brain tumor detection is challenging due to variations in its size, location, and similarity in health tissues. Further, early detection of brain tumors plays a vital role in enabling a wider range of treatment options, improvements in survival rate, and patients quality of life with reduced healthcare costs. That’s why there is a need for an automatic, accurate system that can detect brain tumor at an earlier stage with an accurate result. So, in this research article, pretrained modals such as AlexNet, VGG16, VGG19, ResNet50, InceptionV3, DenseNet121, and all variants of the EfficientNet model were chosen because of their exceptional performance in tasks like feature extraction and image classification with the capability of detecting anomalies in images. The experiment was done on a publicly available multiclass dataset for brain tumor. Further, before passing the dataset to the model, preprocessing was applied to the dataset in terms of resizing the images to the same size; further images were cropped due to the extra boundaries around the images; noise was removed from the images using the FastNIMeans Denoising colored filter; and a data augmentation technique was applied to reduce the overfitting problem and ensure accurate fast training. On different pretrained deep learning models, the EfficientNetB7 model produced better results in terms of validation accuracy. Further pretrained EfficientNetB7 model were customized by adding a few layers and fine-tuning parameters. Our customized pretrained EfficientNetB7 (CPEB7) model was evaluated in terms of accuracy, loss, precision, sensitivity, specificity, recall, f1-score, and MIOU (mean intersection over union) and the model achieved accuracy of 98.57%, 98.97%, 98.97%, and 99.38% for no, meningioma, pituitary, and glioma tumor classes, respectively. Further, our proposed model achieved an overall accuracy of 98.97% with a 1.02% miss classification rate and MIOU of 95.73%. Further, the proposed model was also evaluated on k-fold cross-validation and achieved 99.097% accuracy on fold-5. This proposed system shows superiority over already existing methods in terms of performance evaluation parameters.
- 19 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 11, 117210-117230
Open AccessImproved Multiclass Brain Tumor Detection via Customized Pretrained EfficientNetB7 Model
A brain tumor considered the deadliest disease in the world. Patients with misdiagnoses and insufficient treatment have a lower chance of surviving for life. However, for diagnosing the disturbance in the brain, magnetic resonance images play a vital role, but due to the large number of images produced by MRI, it is time-consuming and difficult to diagnose a patient at an earlier stage with high accuracy. Brain tumor detection is challenging due to variations in its size, location, and similarity in health tissues. Further, early detection of brain tumors plays a vital role in enabling a wider range of treatment options, improvements in survival rate, and patients quality of life with reduced healthcare costs. That’s why there is a need for an automatic, accurate system that can detect brain tumor at an earlier stage with an accurate result. So, in this research article, pretrained modals such as AlexNet, VGG16, VGG19, ResNet50, InceptionV3, DenseNet121, and all variants of the EfficientNet model were chosen because of their exceptional performance in tasks like feature extraction and image classification with the capability of detecting anomalies in images. The experiment was done on a publicly available multiclass dataset for brain tumor. Further, before passing the dataset to the model, preprocessing was applied to the dataset in terms of resizing the images to the same size; further images were cropped due to the extra boundaries around the images; noise was removed from the images using the FastNIMeans Denoising colored filter; and a data augmentation technique was applied to reduce the overfitting problem and ensure accurate fast training. On different pretrained deep learning models, the EfficientNetB7 model produced better results in terms of validation accuracy. Further pretrained EfficientNetB7 model were customized by adding a few layers and fine-tuning parameters. Our customized pretrained EfficientNetB7 (CPEB7) model was evaluated in terms of accuracy, loss, precision, sensitivity, specificity, recall, f1-score, and MIOU (mean intersection over union) and the model achieved accuracy of 98.57%, 98.97%, 98.97%, and 99.38% for no, meningioma, pituitary, and glioma tumor classes, respectively. Further, our proposed model achieved an overall accuracy of 98.97% with a 1.02% miss classification rate and MIOU of 95.73%. Further, the proposed model was also evaluated on k-fold cross-validation and achieved 99.097% accuracy on fold-5. This proposed system shows superiority over already existing methods in terms of performance evaluation parameters.
- 18 August 2023
- Published by Wiley in International Journal of Imaging Systems and Technology
- Vol. 34 (1)
Performance analysis of state‐of‐the‐art CNN architectures for brain tumour detection
Deep learning models, such as convolutional neural network (CNN), are popular now a day to solve various complex problems in medical and other fields, such as image classification, object detection, recommendation of images, processing of natural languages and video and image analysis. So, the idea of studying the architecture of CNNs has gotten a lot of attention and become popular. This study analysed and contrasted the performance of many different CNN models trained on the publicly accessible Br35h dataset for the detection of brain tumours. These models included the LeNet, AlexNet, VGG16, VGG19 and ResNet50. Several optimisers were used in this research to fine‐tune the performance of the CNN model. These included Adam (adaptive moment estimation), SGD (stochastic gradient descent) and RMSprop (root‐mean‐square propagation). Accuracy, miss‐classification rate, sensitivity, specificity, NPV (negative predictive value), PPV (positive predictive value), F1‐score and false omission rate (FOR) were used to assess the efficacy of five different CNN architectures trained using three different optimisers. The experimental results showed that AlexNet architecture with SGD optimiser performed better than other CNN architecture with different optimisers and achieved the highest accuracy of 98.79% with a miss classification rate of 1.20%. It also achieved 98.98% sensitivity, 98.58% specificity, 98.93% NPV, 98.65% PPV, 98.82% F1‐score and 1.06% FOR.
- 18 August 2023
- Published by Wiley in International Journal of Imaging Systems and Technology
- Vol. 34 (1)
Performance analysis of state‐of‐the‐art CNN architectures for brain tumour detection
Deep learning models, such as convolutional neural network (CNN), are popular now a day to solve various complex problems in medical and other fields, such as image classification, object detection, recommendation of images, processing of natural languages and video and image analysis. So, the idea of studying the architecture of CNNs has gotten a lot of attention and become popular. This study analysed and contrasted the performance of many different CNN models trained on the publicly accessible Br35h dataset for the detection of brain tumours. These models included the LeNet, AlexNet, VGG16, VGG19 and ResNet50. Several optimisers were used in this research to fine‐tune the performance of the CNN model. These included Adam (adaptive moment estimation), SGD (stochastic gradient descent) and RMSprop (root‐mean‐square propagation). Accuracy, miss‐classification rate, sensitivity, specificity, NPV (negative predictive value), PPV (positive predictive value), F1‐score and false omission rate (FOR) were used to assess the efficacy of five different CNN architectures trained using three different optimisers. The experimental results showed that AlexNet architecture with SGD optimiser performed better than other CNN architecture with different optimisers and achieved the highest accuracy of 98.79% with a miss classification rate of 1.20%. It also achieved 98.98% sensitivity, 98.58% specificity, 98.93% NPV, 98.65% PPV, 98.82% F1‐score and 1.06% FOR.
- 1 January 2021
- Published by The World Academy of Research in Science and Engineering in International Journal of Advanced Trends in Computer Science and Engineering
- Vol. 10 (3) , 2479-2494
Intelligent Hybrid Fraud Detection Using Biometric and Face Recognition
- 1 January 2021
- Published by The World Academy of Research in Science and Engineering in International Journal of Advanced Trends in Computer Science and Engineering
- Vol. 10 (3) , 2479-2494
Intelligent Hybrid Fraud Detection Using Biometric and Face Recognition