NUST Institutions Library Catalogue Search for 'an:"119662"'

Automated Segmentation and Classification of Lesion On Breast Ultrasound

By Zahra Fazal. Islamabad SMME NUST 2014 . 56 p. 30cm.

Place Hold on Automated Segmentation and Classification of Lesion On Breast Ultrasound

White Matter Multiple Sclerosis Lesion Segmentation Under Distributional Shifts /

By Haider, Ali . . 37p. ; 30cm..

Place Hold on White Matter Multiple Sclerosis Lesion Segmentation Under Distributional Shifts /

Vehicle Detection and Tracking in Aerial Imagery /

By Tahir, Muhammd Abdullah . . 42p. 30cm.

Place Hold on Vehicle Detection and Tracking in Aerial Imagery /

Deploying Efficient Net (BNs) for grading Diabetic Retinopathy severity levels from fundus images /

By Batool, Summiya . . 40p. , One of the common and escalating endocrine illnesses is diabetes mellitus. Diabetic retinopathy is a common eye problem in patients with diabetes. Diabetic retinopathy (DR), a retinal condition, is acknowledged as an epidemic on a global scale. One-third of the estimated 285 million persons with diabetes show symptoms of DR, and one-third of them have DR that threatens their vision [1]. In addition, the figures are rising. By 2040, 288 million individuals are expected to have AMD, and by 2050, the number of people with DR is projected to treble. The need for reliable diabetic retinopathy screening systems became a critical issue recently due to the increase in the number of diabetic patients. The severity of DR can be graded into five stages: normal, mild NPDR, moderate NPDR, severe NPDR, and PDR. Early diagnosis and treatment of DR can be accomplished by organizing large regular screening programs. Numerous Convolutional neural network models are developed for the diagnosis of DR in fundus images using deep learning methods. In Deep Learning (DL) one of the methods is a computer-aided medical diagnosis for the detection of DR. There are many DL-based methods such as restricted Boltzmann Machines, convolutional neural networks (CNNs), auto-encoder, and sparse coding. On the other hand, it is thought-provoking to distinguish it initially not display signs in the initial classes. The current models for diabetic retinopathy mayn’t identify entire classes of DR. The utmost commonly used metrics like accuracy, f1-score, precision, and recall; do not cogitate the standard of difference among labels, which supports spotting all classes of DR. In our paper used Efficient Net BNs models. We concluded evaluation scores using the F1-score, which is applicable for grading various classes of DR established on the intensity levels. We have accomplished the F1- score of 0.88 and 0.84 using the simple preprocess, Gaussian smoothing filters, and deploying an Efficient Net BNs network on DeepDRiD and EYE-PACS datasets. 30cm.

Place Hold on Deploying Efficient Net (BNs) for grading Diabetic Retinopathy severity levels from fundus images /

Bone X-ray abnormality detection using MURA dataset /

By Batool, Sana. . 40p. , Musculoskeletal abnormalities along with bone fractures are a wide range of abnormalities that account for most visits of patients to Emergency department of hospitals. According to an estimate, more than 1.7 billion people are affected by musculoskeletal disorders each year. Bone X-rays are the first line imaging modality for imaging of fractured bones. Radiologists then undergo reporting of X-rays for detection of fractures and pathologies. Classification of bone X-rays into normal and abnormal is a time-taking process and is also subjected to variability between different radiologists. Therefore, the use of automatic classifiers incorporating deep learning algorithms is currently in use in clinical diagnostics. MURA is a large publicly available dataset released by the machine learning group of Stanford university. MURA dataset consists of 40,895 multi-view images of upper limb that belong to seven regions namely shoulder, humerus, elbow, forearm, wrist, hand, and fingers. In this study we propose the use of the single DenseNet-169 model trained on complete dataset along with multiple preprocessing and data augmentation steps, based on Keras in TensorFlow. Training data was divided into 80:20 for training and validation respectively, whereas, testing of model was done on validation set. The results obtained through the proposed technique include 80% testing accuracy. This validates the effectiveness of this method for bone fractures classification. 30cm.

Place Hold on Bone X-ray abnormality detection using MURA dataset /

Automated Brain Tumor Segmentation using Multimodal MRI Scans /

By Ehsan ,Fatima . . 73p. ; 30cm..

Place Hold on Automated Brain Tumor Segmentation using Multimodal MRI Scans /

EYE MOVEMENT BEHAVIOUR VARIES WITH EMOTIONAL CUES /

By KHALID, USMAN . . 69p. ; 33cm..

Place Hold on EYE MOVEMENT BEHAVIOUR VARIES WITH EMOTIONAL CUES /

Melanoma Detection Using Machine Learning /

By Zafar ,Kashan . . 70p. ; 30cm..

Place Hold on Melanoma Detection Using Machine Learning /

Multilabel Classification and Localization of Rare Pulmonary Diseases using Deep Learning /

By Khaliq, Fariha . . 52p. , Chest radiography is the most common radiological examination used for the diagnosis of thoracic diseases. Currently, automated classification of radiological images is abundantly used in clinical diagnosis. However, each pathology has its own response characteristic receptive field regions, which is a key problem during the classification of chest diseases. In addition to extreme class imbalance, cases labelled as uncertain in the dataset further complicate this task. To solve this problem, we propose a semi-supervised learning approach known as U-SelfTrained. In this scheme, uncertain labels are left unlabeled in the dataset; first, the model is trained on labelled instances and then on unlabeled instances relabeling them with labels having a higher probability. Comprehensive experimentation was carried out on the CheXpert dataset, which consists of 223,816 frontal and lateral view CXR images of 64,740 patients with 14 diseases. The testing accuracy is 0.877 on the CheXpert dataset, which is currently the highest score achieved to date. This validates the effectiveness of this method for chest radiography classification. The practical significance of this study is the implementation of AI algorithms to assist radiologists in improving their diagnostic accuracy. 30cm.

Place Hold on Multilabel Classification and Localization of Rare Pulmonary Diseases using Deep Learning /

Time-Frequency Analysis of Fatigue in Lower Limb Muscles Under External Super- Imposed Vibrations /

By Ashfaque ,Moeez. . 52P. ; 30cm..

Place Hold on Time-Frequency Analysis of Fatigue in Lower Limb Muscles Under External Super- Imposed Vibrations /

Liver Segmentation from Combined Tomography (CT) Abdominal Images Data Using Deep Neural Networks /

By tuz Zahra Khan ,Fatima . . 60p. ; 30cm..

Place Hold on Liver Segmentation from Combined Tomography (CT) Abdominal Images Data Using Deep Neural Networks /

Multi-Disease Classification For Retinal Diseases Using Deep Learning Technique

By Khan, Omar Salman . . 55p. , Diagnosis before the spread of retinal diseases is vital to prevent high level blindness and any sort of visual impairment. Many retinal diseases can be found via the fundus imaging which has a very important role in the observation and detection of various ophthalmic diseases. Most previous literature has focused their approaches on identifying individual diseases or a combination of 3-4 diseases like DR, MYA, ARMD, MH, ODC having major research. The eye is mostly affected by more than one underlying disease or disease marker, and uptil now most datasets had very few classes. Recently introduced RFMiD dataset, is one of the first datasets to provide 45 different classes of ophthalmic diseases. Hence making it possible to work towards automated multi-disease classification models which would provide great help to highlight this issue via clinical decision support systems integrated in the medical image diagnosis. Our work aimed to achieve higher accuracy than previous literature and to create an CDS application from the model in understanding and predicting multi retinal diseases. Deep learning models are excellent and have proven to be extremely effective in solving complex image processing problems. In addition, ensemble learning yields high generalization performance by reducing variance. Therefore, a synthesis of transfer, ensemble, and deep learning was used in this work to create an accurate and reliable model for multi retinal disease classification. To create the Multi Retinal Disease Classification Model (MRDCM) we used ensemble of EfficientNetB4 and EfficientNetV2S, with our final ensemble model giving promising results. In our evaluation, we scored an AUC of 0.973 which stands better than literature. Further our model selection is lighter than models used in literature. The model was tested on 27 main classes of RFMiD dataset for comparison with literature. Index Terms—Deep iii Learning, Ensemble learning, Retinal Image Analysis, multi-Disease classification, transfer learning. 30cm.

Place Hold on Multi-Disease Classification For Retinal Diseases Using Deep Learning Technique

Ocular Disease Intelligent Recognition /

By Sahar, Syeda Ghina . . 46p. , To record anatomical details of the eye and anomalies, fundus imaging has proved very efficient. The most effective way to see and diagnose a wide range of eye diseases is through fundus imaging. Conditions that affect the blood vessels and areas surrounding it include diabetesrelated retinopathy, glaucoma, AMD, myopia, cataract and hypertension. It's possible for the patient to have more than one ophthalmological problems that can be seen in one or both of his eyes. The dataset provided by ODIR is used in this study. The data has eight different categories for the diseases to be detected. By using transfer learning, two simultaneous models are described for solving the multi label problem for both the eyes (left and right). For the convolutional network, two synchronous efficient net models are implemented which are used with ADAM optimizers for better detection and results outcome. On the ODIR data set, B7 Efficient net along with focal loss outperformed the other approaches with an accuracy rate of 0.96%. 30cm.

Place Hold on Ocular Disease Intelligent Recognition /

Kidney and Kidney Tumor Segmentation, 2019 (KiTS-19) /

By Abbasi, Ramsha . . 58p. , Computed Tomography (CT) is the most widely used imaging procedure for locating and diagnosing kidney tumors. The standard treatment for kidney tumors is surgical removal. It is important to accurately segment the kidney and its tumor for effective surgical planning. The manual segmentation of kidney tumors is time-consuming and subject to variability between different radiologists. Therefore, automatic semantic segmentation of kidney tumors using deep learning networks has become increasingly popular in the past few years. Automatic segmentation of kidney tumors is a very challenging task due to their morphological heterogenicity. This work provides the application of 3D UNet and 3D SegResNet on KiTS19 challenge data for accurate segmentation of kidney and kidney tumors. An ensembling operation was added in the end to average the predictions of all models. The proposed method is based on the MONAI framework and focuses more on training procedure rather than complex architectural modifications. The models were trained using KiTS19 training set of 210 cases for which ground truth labels were available. The training data was divided into 190:20, for training and validation respectively. We evaluated the performance of our network on KiTS19 official test set and obtained mean dice of 0.8964, 0.9724 kidney dice, and 0.8204. Our approach outperforms many submissions in terms of kidney segmentation and gives promising results for tumor segmentation. We also used a local test set of 90 cases from KiTS21 challenge to check how well our method adopts to a new dataset. It scored a mean dice of 0.9160, kidney dice of 0.9771, and 0.8550 tumor dice. The obtained results on KiTS19 official test set and local test set show that our approach is effective and can be used for organ segmentation. 30cm.

Place Hold on Kidney and Kidney Tumor Segmentation, 2019 (KiTS-19) /

3D Neural Network for Detection of ACL Injury in Knee MRI Scans /

By Kamran, Abdullah . . 43p. , Computer aided diagnosis is widely used in medical imaging for the diagnosis of many diseases such as cardiomegaly, brain and kidney tumor, lung cancer, COVID-19 and may more. For the past few decades, computer aided diagnosis has significantly improved due to the development of better architecture used for the diagnosis. Knee injury diagnosis using deep learning techniques is highly popular due its high detection rate and is highly localized. Many state-of-the-art-deep learning models have been used for the detection of abnormalities, meniscus tear and ACL tears in Knee MRI scans. These models include RESNET, Google-Net, VGG19 and VGG16, Alex-Net and many other, all giving significant results. In this study we used a custom 3D CNN model which is light in weight. For training we are using two datasets, one provided by Stanford ML group and the other form Hospital in Croatia. We combined the two dataset and split it into 80-20 ration (80% of the data used for training and remaining for testing purposes). Both the dataset has extreme class imbalance, so we used data augmentation and class weights to rectify its effect on the training process. Further the voxel intensities for the two datasets were different (one dataset was in 8-bit format and the second was in 12-bit format), so we normalized the intensity values using mathematical formulas. For contrast, we performed adaptive histogram equalization Average accuracy and AUC achieved by our model on training set is 97.6 and 99.3 respectively, during 5-fold cross validation. 30cm.

Place Hold on 3D Neural Network for Detection of ACL Injury in Knee MRI Scans /

Facial Micro-Expression Detection under Variable Road Condition for Evaluation of Driver’s Performance /

By Shaikh, Aleema Moin . . 52p. , The emotional state of the driver has a direct impact on driving. Hence, to ensure road safety, it is important to monitor driver’s emotional state. This challenging task can be achieved through analyzing micro-expressions as they are linked with a person’s emotional state and emerge on face even under situations where a person is trying to conceal true emotions. Moreover, it’s easy to acquire facial data while driving than any other stress signal. This research focuses on identifying micro-expressions linked with stressful emotional state in drivers. The physiological parameters like heart rate and stress value based on heart rate variability are also monitored as they fluctuate easily under emotional changes within the body. This research considered the emotions of happiness, sadness, surprise, anger, fear and disgust. To evaluate stress within drivers, the dominant emotion behind detected micro-expression is found through an emotion detection open source code. The results show a high F1 score for the identified micro-expressions i.e. 1.00, 0.947, 0.933 and 0.85. These findings can help in face readings where stress detection is required and can contribute towards better systems in cars to ensure road safety and manage stress. 30cm.

Place Hold on Facial Micro-Expression Detection under Variable Road Condition for Evaluation of Driver’s Performance /

Detection of COVID’19 through low resolution CT scan images using Deep Learning /

By Arif, Hajra . . 51p , Coronavirus emerged as a deadly disease in 2019 killing almost 6.25 million people, with its variants still being discovered. For timely medical treatment it is necessary to accurately and rapidly diagnose the disease. The main test used for diagnosis was “The Reverse Transcription Polymerase Chain Reaction (RT-PCR)” but due to limited availability of RT-PCR equipment at the time of outbreak, alternative methods were used to mitigate the damage. One of them was the computed tomography (CT) scans, a noninvasive imaging approach. Utilizing this CT data, deep learning (DL) models were developed to expedite the diagnostic procedure. Due to privacy concerns, the original CT scans were not shared with the public which caused hindrance in the research and development of accurate DL methods. Hence, datasets were made from secondary sources, either by extracting images from preprints or saving images in any other format than DICOM, which generated low resolution images. To address this issue, preprocessing techniques were applied to generate better results of the DL models. The pixel intensities in images are normalized such that they lie in the range of the actual values of a CT scan in the Hounsfield unit scale and then given as an input to the model. Diagnostic performance was assessed by F1-score (84%), AUC (94%) and Accuracy (81%), which is better than the performance achieved without pre-processing. This study proves that enhancing the image quality, through pre-processing techniques, can improve the results when good quality data is unavailable and accurate models can be made for detecting any disease at the time of the outbreak. 30cm.

Place Hold on Detection of COVID’19 through low resolution CT scan images using Deep Learning /

Deep Learning for Improved Myoelectric Control /

By Zia ur Rehman, Muhammad . . 153p. , Advancement in the myoelectric interfaces have increased the use of myoelectric controlled robotic arms for partial-hand amputees as compared to body-powered arms. Current clinical approaches based on conventional (on/off and direct) control are limited to few degree of freedom (DoF) movements which are being better addressed with pattern recognition (PR) based control schemes. Performance of any PR based scheme heavily relies on optimal features set. Although, such schemes have shown to be very effective in short-term laboratory recordings, but they are limited by unsatisfactory robustness to non-stationarities (e.g. changes in electrode positions and skin-electrode interface). Moreover, electromyographic (EMG) signals are stochastic in nature and recent studies have shown that their classification accuracies vary significantly over time. Hence, the key challenge is not the laboratory shortterm conditions but the daily use. Thus, this work makes use of the longitudinal approaches with deep learning in comparison to classical machine learning techniques to myoelectric control and explores the real potential of both surface and intramuscular EMG in classifying different hand movements recorded over multiple days. To the best of our knowledge, for the first time, it also explores the feasibility of using raw (bipolar) EMG as input to deep networks. Task are completed with two different studies that were performed with different datasets. In the first study, surface and intramuscular EMG data of eleven wrist movements were recorded concurrently over six channels (each) from ten able-bodied and six amputee subjects for consecutive seven days. Performance of stacked sparse autoencoders (SSAE), an emerging deep learning technique, was evaluated in comparison with state of art LDA using offline xii classification error as performance matric. Further, performance of surface and intramuscular EMG was also compared with respect to time. Results of different analyses showed that SSAE outperformed LDA. Although there was no significant difference found between surface and intramuscular EMG in within day analysis but surface EMG significantly outperformed intramuscular EMG in long-term assessment. In the second study, surface EMG data of seven able-bodied were recorded over eight channels using Myo armband (wearable EMG sensors). The protocol was set such that each subject performed seven movements with ten repetitions per session. Data was recorded for consecutive fifteen days with two sessions per day. Performance of convolutional neural network (CNN with raw EMG), SSAE (both with raw data and features) and LDA were evaluated offline using classification error as performance matric. Results of both the short and long-term analyses showed that CNN and SSAE-f outperformed the others while there was no difference found between the two. Overall, this dissertation concludes that deep learning techniques are promising approaches in improving myoelectric control schemes. SSAE generalizes well with hand-crafted features but fails to generalize with raw data. CNN based approach is more promising as it achieved optimal performance without the need to select featur 30cm..

Place Hold on Deep Learning for Improved Myoelectric Control /

DETECTION OF THYROID DISEASES USING MACHINE LEARNING TECHNIQUES /

By AKHTAR,TEHSEEN . . 137p. , Background: The unusual growth of the glandular tissue on the boundary of the Thyroid gland is an indication of Thyroid disease. Thyroid disease is characterised by an unusually high or low number of hormones produced by the thyroid gland, the two most prevalent kinds are hypothyroidism (underactive thyroid gland) and hyperthyroidism (overactive thyroid gland). The main aim of this project was to introduce the concept of an efficient multi-stage ensemble i.e., the voting ensemble of the homogeneous ensemble which could be used with a variety of feature-selection algorithms for improving the diagnosis of thyroid diseases. The dataset utilised in this study was built from real-time thyroid data obtained from the teaching hospital in DG Khan at District Head Quarter (DHQ), Pakistan. Following the appropriate preprocessing processes, three kinds of attribute-selection strategies were used: The first approach used was Select from Model (SFM), the second technique was the Select K-Best (SKB), and the final methodology was the Recursive Feature Elimination (RFE). Select From Model (SFM) is a form of attribute-selection strategy that uses a model to select attributes. As potential feature estimators, the Decision Tree (DT), Logistic Regression (LR), Gradient Boosting (GB) and Random Forest denoted as the (RF) classifiers were employed in conjunction with each other. The homogeneous ensemble activated the bagging, boosting-based learners, who were then classified by the Voting ensemble, which employed both soft and hard voting to categorise the data. Other performance assessment criteria such as hamming loss, accuracy, mean square error, sensitivity and others have been implemented. The results of the experiments reveal that when the suggested approach for better thyroid sickness detection is applied in its most practicable form, it is most successful. On the dataset 1, all of the algorithms tested obtained 100 % accuracy with subset of the total no of feature in each case, however on the dataset 2, more than 98 percent accuracy was reached in every case. On the basis of accuracy and computing cost, the results given here exceeded equivalent benchmark models in their respective fields of study. 30cm.

Place Hold on DETECTION OF THYROID DISEASES USING MACHINE LEARNING TECHNIQUES /