PhD

Now showing 1 - 2 of 2

Automatic Screening of Diabetic Retinopathy at Early Stage using Colored Fundus Images
(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Muhammad Nadeem Ashraf; FA14-PCS-003; LHR TP 7864; Dr. Zulfiqar Habib
The research work in this thesis is about an automated diagnosis of diabetic retinopathy (DR) for screening of DR patients at an early stage by classifying both red lesions using color fundus images, through a deep convolutional neural network (DCNN). DR is an asymptotic and vision-threatening retinal complication among working-age adults. Computer-aided diagnosis (CAD) is useful in screening DR patients to assist ophthalmologists to prevent blindness. Researchers have focused on this area by proposing many automated systems to diagnose DR. A comprehensive literature search is conducted during this research. Early-stage diagnosis is challenging due to poor representation of less discriminative and small-sized red lesions of DR. Traditional handcrafted based methods are not usually recommended for clinical trials due to their various limitations. Modern DCNNs are popular to solve various computer vision problems accurately. However, the requirement of huge training data restricts training deep models from scratch to solve the problems in medical imaging. Transfer learning and fine-tuning are helpful alternatives, but overfitting and poor performance further demand the architectural amendments for effective use of the deep CNN models in medical imaging with smaller datasets. Various pre-trained CNN models (AlexNet, VGG16, GoogLeNet, Inception-v3, ResNet50, and DenseNet) are fine-tuned, on the augmented set of 200 × 200 small image patches from the lesion-level annotated images of e-Ophtha_MA dataset, through transfer learning and hyper-parameter tuning to analyze their performance to classify both red lesions of DR. The architecture of a high-performing deep ResNet50 model is selected to fine-tune further. Its baseline architecture is modified by introducing; i) the reinforced skip connections, ii) a Global Max pooling layer, and iii) the Sum-of-Squared Error (SSE) Loss function. The suggested modifications are robust and straightforward. Such alterations may help to enhance various CNN architectures to apply in other imaging domains with small datasets. The proposed framework is evaluated on five publically available datasets: e Ophtha_MA, DiaRetDB1 v2.1, ROC, IDRiD, and Messidor by computing many xi performance metrics. The highest scores (0.9851, 0.991, 0.991, 0.991, 0.991, 0.9939, 0.0029, 0.9879, and 0.9879) of the sensitivity, specificity, AUC, accuracy, precision, F1-score, false-positive rate, Matthews’s correlation coefficient, and kappa coefficient are obtained on unseen test instances from e-Ophtha_MA for DR detection, respectively. The cross-validation results obtained by the proposed method on Messidor and IDRiD datasets at the image level are also better than state-of-the-art techniques. The achieved results are promising and demonstrate the effectiveness of a suggested architecture. For qualitative assessments, the gradient of class activations mapping (Grad-CAM) is computed to visualize the decision for each classified instance. This visual interpretation gives more satisfaction to using the proposed framework for clinical validations. Due to performance, simplicity, and robustness, the suggested model is suitable for diagnosing DR at an early stage for the screening of DR patients. It would help to develop a reliable health care system.
Robust Image Forgery Detection and Localization using Texture Information
(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2018) Muhammad Khurshid Asghar; SP14-PCS-002; LHR TP 5604; Dr. Zulfiqar Habib
Forensic analysis of images has important applications in security, court of law, insurance agencies, medical imaging, and electronic media. To develop robust image forgery detection and localization methods, applicable to real scenarios are highly needed. The focus of this thesis is to develop automatic image forgery detection and localization techniques. In the image forgery detection and localization, the main challenge is the robust representation of tampering traces. First, to identify the research gaps for contribution, the state-of-the-art passive image forgery detection and localization techniques have been reviewed. The existing techniques are not robust in representing the tampering traces. To overcome this issue two techniques have been proposed. The first technique has been developed to detect copy-move forgery in images by estimating noise patterns. To represent noise patterns, a new descriptor FFT-DRLBP employing Fast Fourier Transformation (FFT) and Discriminative Robust Local Binary Patterns (DRLBP) is introduced. Noise patterns are estimated using FFT, then the discrepancies in the noise patterns due to tampering are encoded using DRLBP. Support Vector Machine (SVM) is used to classify images as authentic or forged. This technique detects authentic and forged images with 99.21 % accuracy. To localize the copy-move forgeries a robust FFT-SIFT descriptor based on FFT and Scale Invariant Feature Transform (SIFT) is proposed. Localization method based on FFT-SIFT descriptor outperforms state-of-the-art and achieves high true positive rate while maintaining low false positive rate. The second technique has been developed to detect splicing forgery in images by estimating noise inconsistencies. For this purpose, a new descriptor DWT-DRLBP is introduced based on Discrete Wavelet Transformation (DWT) and DRLBP. First image is decomposed using DWT, the texture variation in each DWT sub-band is encoded using DRLBP histograms. Cb and Cr components are used to extract features using DWT-DRLBP descriptor. For classification SVM is employed. The method offers excellent results (98.95 %) and outperforms the state-of-the-art methods. xi In addition, to validate the robustness of the proposed methods on real scenarios, a new dataset called Forged Real Images Throughout History (FRITH) is developed to validate the performance of the proposed methods. To further validate the robustness of the proposed methods, cross-dataset experiments are performed to analyze the applicability on unseen images

PhD

Browse

Filters

Settings

Sort By

Results per page

Search Results