DSpace Repository

Nazarbayev University Repository (NUR) is an institutional electronic archive designed for the long-term preservation, aggregation, and dissemination of scientific research outcomes and intellectual property produced by the Nazarbayev University community and affiliated organizations.

Communities in DSpace

Select a community to browse its collections.

Now showing 1 - 5 of 5

001. Theses 3085
002. NU Research Publications 11760
003. NU Research Institutes & Centres 586
004. NU Journals & Magazines 285
005. Forums, Conferences & Projects 1332

Recent Submissions

Access status: Embargo ,
SarcomaNet: A Privacy-Preserving Multi-Task Deep Learning Framework for Soft Tissue Sarcoma Analysis from Multi-Modal Imaging
(Nazarbayev University School of Engineering and Digital Sciences, 2026-05-04) Rasool, Muhammad Husnain; Rai, Hari Mohan
Soft tissue sarcomas (STSs) are a rare and biologically heterogeneous group of malignancies comprising more than 100 histological subtypes, with an annual incidence of fewer than 5 per 100,000 persons. Their rarity makes large, well-annotated imaging datasets difficult to assemble, limiting the use of deep learning for automated tumour segmentation, histological grade classification, and survival risk estimation—three clinically relevant tasks that are often performed separately and with limited quantitative imaging support. This thesis presents SarcomaNet, a unified, privacy preserving tri-head deep learning framework that addresses all three tasks jointly from multi-modal imaging using the publicly available TCIA Soft Tissue Sarcoma dataset (𝑁 = 51). SarcomaNet is built on a shared 3D Residual U-Net (ResU-Net) encoder–decoder backbone and introduces four complementary innovations to address the constraints of rare-cancer imaging. (1) Cross-Modal Masked Autoencoding (CrMAE) is a self-supervised pre-training strategy that leverages the complementary biophysics of multi-modal imaging by masking a randomly selected modality channel (T2FS MRI, FDG-PET, or CT) at 50% patch density and training the encoder to reconstruct the masked modality from the two visible channels. Unlike spatial masked autoencoders, CrMAE encourages the encoder to learn physics-grounded cross-modal correspondences (oedema ↔ metabolic activity ↔ tissue ensity) hat are directly informative of tumour biology. Over 40 pre-training epochs on all 51 unlabelled patient volumes, CrMAE reduces reconstruction MSE by 87.7% without requiring external data...
Access status: Open Access ,
A Video-Based Reverse Dictionary for Sign Language Using Gesture Similarity
(Nazarbayev University School of Engineering and Digital Sciences, 2026-05-04) Orazumbekov, Batyrbek; Sandygulova, Anara
Most sign language recognition systems have been developed as classification models that associate gesture videos with pre-defined glosses, but such systems do not facilitate similarity search, where users can make queries without knowing the labels of gestures. This thesis proposes a sign language retrieval system based on pose representation that functions as a reverse gesture dictionary, enabling users to retrieve visually similar gestures directly from video input. The proposed method converts gestures into normalized skeletal joints rather than RGB images to minimize variations in appearance, such as background, lighting, and clothing, and to focus on dynamic motion patterns. The extracted keypoints are temporally normalized and optionally augmented with motion features to better capture gesture dynamics. In order to model the temporal relationships within the data, two models are considered; one being a Transformer model with a self-attention mechanism and another one being a Spatial-Temporal Graph Convolutional Network (ST-GCN). Both of these can be used to compare the capabilities of sequence models in modeling temporal dependencies. The model was evaluated using the WLASL dataset under the few-shot setting and ranking metrics like Recall@K and mean Average Precision (mAP), rather than using classification accuracy as it better suits a retrieval task. According to experimental results, it can be concluded that the Transformer model performs better when it comes to modeling temporal relationships between frames in gesture sequences compared to graph-based models. Additionally, employing attention-driven pooling during temporal aggregation improves the results significantly and achieves an mAP of 0.237 on the validation set. Transferability of the embedding space to novel gestures is tested by applying the trained model to the AUTSL dataset (using only a subset of 226 labels). Finally, the impact of approximate nearest neighbor search on retrieval results is examined.
Access status: Open Access ,
Design of Graphene Membrane For Wastewater Treatment
(Nazarbayev University School of Engineering and Digital Sciences, 2026-05-26) Uma-Oji, Stephen ; Kostas, Konstantinos; Sarbassov, Yerbol
The growing global need for fresh water has driven the development of advanced membrane technologies with not only high-water permeability but also good rejection of contaminants and long-term operational stability. Graphene-based membranes have been considered as the promising candidate for next-generation water purification because of its atomic thickness, high mechanical strength and tunable nanoporous structure. We present here a study of the performance of various functionalized graphene nanopore membranes in water treatment via MD simulations. Two broad strategies for functionalization were explored: uniform surface functionalization, and pore-edge functionalization. Three different chemical groups, hydrogen (H), fluorine (F), and amine (NH₂) were chemically functionalized on graphene membranes at different levels of functionalization. Functionalization was implemented uniformly over the surface at a coverage of 20%, and pore-specific functionalization localized to regions around the edges of the nanopore at a coverage of 15%. The simulation system includes the pressure-driven water through graphene nanopores helping remove of contaminants such as PFBA⁻ anions and Pb²⁺ ions. Molecular dynamics simulations were carried out using the LAMMPS simulation package with explicit water models and realistic intermolecular interaction parameters. To accelerate transport processes, they were subjected to a piston-driven pressure of 20,000 bar and the resulting flux values were scaled to realistic operating pressures of 100 bar. These results show that pore chemistry is a key parameter in determining membrane transport behavior. Also, Fluorinated graphene membranes achieved the highest water flow rates since their reduced friction and hydrophobicity of channel structure favor almost frictionless transport of water in the nanopore. The permeability was a bit lower for hydrogen-functionalized membranes, but the transport properties remained stable. Fouling index analysis suggested that pore functionalization could generally reduce fouling relative to uniform functionalization owing to localization of chemical groups in proximity to the nanopore region which mitigates surface-wide contaminant adsorption. Electrostatic interactions occurring between the functional groups and charged species such as heavy metal ions and PFAS compounds are the primary driving mechanism for selectivity. Pore-functionalized membranes exhibited the best balance of desirable characteristics (high water flux, strong rejection, reduced foulant deposition tendency) among the studied systems.
Access status: Embargo ,
Spatial Modeling for Debris Flow Analysis
(Nazarbayev University School of Engineering and Digital Sciences, 2026) Tanzhanova, Luiza; Kim, Jong; Satyanaga, Alfrendo
Debris flows are among the most hazardous natural phenomena in mountainous regions, often resulting in severe infrastructure damage and loss of life. The study examines the physical and rheological characteristics of debris-flow material obtained in the area of a dam that traps debris in Almaty, Kazakhstan. To identify the grain size distribution, atterberg limits, specific gravity, viscosity, and yield stress, laboratory tests were carried out. The laboratory results indicated that the investigated material is predominantly coarse-grained with a limited proportion of fine particles. Rheological analysis demonstrated behaviour consistent with a Bingham-type fluid, where both yield stress and viscosity decreased with increasing water content. Numerical analyses using FLO-2D showed that debris-flow propagation in the Kumbelsu River basin was strongly controlled by local topography and material properties, while increasing discharge led to higher maximum flow depth and velocity. The simulations also showed that FLO-2D is useful for representing general flow direction, depth, velocity, and hazard distribution, although discrepancies with the historical event remained because of limitations related to surge waves, DEM differences, and rheological input data.
Access status: Embargo ,
Modified Geogrid for Geobarrier System Incorporating Unsaturated Soil
(Nazarbayev University School of Engineering and Digital Sciences, 2026-04-24) Dewangga, Eriko; Satyanaga, Alfrendo; Kim, Jong Ryeol; Zhang, Dichuan
The Geobarrier System (GBS) is a three-component system to improve slope stability by combining capillary barrier, retaining system and vegetative cover. Currently, the retaining system of the conventional configuration is mainly controlled by the geogrid. However, the embedment length of the geogrid requires massive excavation which also leads to the longer construction duration. Considering this limitation, reducing the length of the geogrid while maintaining sufficient stability became the primary objective of this study. On the other hand, although earth nailing systems have been widely applied in slope stabilization practice, integration between geogrid reinforcement and earth mechanical anchorage has not been demonstrated within a unified framework. Therefore, J-pin attachments to the tip of geogrid are introduced in this study to compensate for the tensile resistance loss of the reduced geogrid embedment length. To conduct this, the conventional model with geogrid embedment length of 0.7 times the slope’s height (H) was modified into 0.6 H, 0.5 H, and 0.4 H. Attachment of J-pins were carried out gradually from the lowest layer of the geogrid. Analysis was conducted with two finite element software: PLAXIS 2D and ABAQUS, each with different purposes. PLAXIS 2D was utilized to perform unsaturated seepage analysis and mechanical analysis for calibration of ABAQUS numerical control settings. ABAQUS was used to perform mechanical analysis calibrated from PLAXIS 2D for modified models with J-pin attachments. Results show that modification on the geogrid embedment length has no significant influence on the pore-water pressure. The mechanical analysis concluded that J-pin attachment only effectively compensates for Model 0.6 H and 0.5 H with the optimum number is five J-pins, confirming that increasing anchorage density cannot compensate for insufficient geogrid embedment length. Furthermore, J-pin contributed up to 19.3% of total reinforcement force demonstrating effective composite interaction with geogrid.