INVITED SPEAKERS

Prof. Xibin Jia, Beijing University of Technology, China

Speech Title: Clinically Oriented Medical Image Analysis

Abstract: Medical imaging plays a fundamental role in modern clinical practice, supporting disease screening, diagnosis, and treatment planning across diverse modalities such as X-ray, MRIs, and CT. Despite significant advances in medical image analysis, two major challenges remain: poor generalization across different clinical environments and a heavy reliance on large-scale annotated datasets. In this talk, I will present our recent work on clinically oriented medical image analysis, focusing on improving model robustness and data efficiency. Specifically, we explore domain generalization, aiming to develop models that can generalize effectively to unseen hospitals or imaging conditions without accessing target-domain data. In addition, we investigate learning with limited annotations, including semi-supervised and few-shot learning approaches, as well as vision–language pre-training, to reduce dependence on costly expert annotations while maintaining high performance.We further introduce our efforts in leveraging emerging architectures, such as Mamba, to capture long-range dependencies while enabling effective local modeling. Finally, we briefly extend our research to causal modeling for complex tasks such as long-term action recognition. Overall, this work aims to bridge the gap between methodological advances and real-world clinical applicability, enabling more effective and efficient medical image analysis.

Biography: Xibin Jia a Professor of College of Computer science at the Beijing University of Technology and Beijing Institute of Artificial Intelligence. She received her Ph.D. in Computer Application Technology from Beijing University of Technology in 2007. She was a one-year visiting scholar at University of California Riverside U.S. and a half-year visiting scholar at Flinders University Australia respectively. She has nearly 30 years of experience in computer science teaching and research. She is a Ph.D. and M.S. supervisor. Her research interests include computer vision, deep representation learning, and multi-modality deep learning. Her current work focuses on Intelligent Medical Image Analysis and Diagnosis, Affective Computing, and behavior recognition. She has served as Principal Investigator (PI) or co-PI on dozens of projects supported by grants from the National Natural Science Foundation of China, the Beijing Natural Science Foundation, and others. She has published dozens of papers in refereed journals and conference proceedings, including ACM MM 2025, MICCAI 2025, IEEE JBHI, ESWA, EAAI, and IEEE/ACM-TCBB. She is a Distinguished Member of CCF and a member of several technical committees, including CCF-CV, CSIG-BVD, CSIG-MV, and CAAI-IM. She has served as an Area Editor and Editor for the SCI-indexed journal KSII-TIIS. Her papers was awarded the IEEE MedAI 2024 Best Paper Award.

 

Prof. Seokwon Yeom, Daegu University, Korea

Speech Title: Thermal Image Detection and Tracking with a Multi-rotor Drone

Abstract: This talk presents thermal image-based object detection and tracking using a multi-rotor drone. Thermal objects are detected by a YOLO detection model trained on a custom dataset, and the detected objects are tracked using Kalman or IMM filters. Through track association and fusion, the most reliable tracks are selected and fused to generate continuous trajectories. In addition, the track segment association connects track segments that are temporally disconnected. In the experiment, three hikers on a mountain were captured using a drone-mounted thermal imaging camera. The proposed approach achieved excellent tracking performance in terms of total track lifetime, mean track lifetime, and track purity.

Biography: Seokwon Yeom has been a faculty member of Daegu University since 2007. He has a Ph.D. in Electrical and Computer Engineering from the University of Connecticut in 2006.
He has been a guest editor of Applied Sciences and Drones in MDPI since 2019. He has served as a board member of the Korean Institute of Intelligent Systems since 2016, and a member of the board of directors of the Korean Institute of Convergence Signal Processing since 2014. He has been program chair of several international conferences. He was a vice director of the AI homecare center and a head of the department of IT convergence engineering at Daegu University in 2020-2023, a visiting scholar at the University of Maryland in 2014, and a director of the Gyeongbuk techno-park specialization center in 2013. He has been a keynote or invited speaker at several international conferences. His research interests are intelligent image and optical information processing, deep and machine learning, target tracking, and state estimation of drones.

 

Prof. Suraiya Jabin, Jamia Millia Islamia, India

Speech Title: AI-based Adult Age Threshold Modeling using Orthopantomographs for Forensic and Judicial Applications

Abstract: Dental development visible in OPG is considered one of the most reliable biological indicators for age estimation for persons. This is the most legally sensitive age group in India: 17 or 18 years can change court jurisdiction, determines juvenile vs adult trial, impacts sentencing severity, etc. Therefore, OPG-based AI-assisted adult/non-adult classification constitutes a scientifically valid, ethically sound, and legally relevant tool to aid forensic age determination in India.
EfficientNet emphasizes efficiency through compound scaling, jointly optimizing network depth, width, and input resolution, resulting in exceptional accuracy; efficiency trade-offs using MBConv blocks and squeeze-and-excitation attention. We fine-tuned these variants of deep convolution networks viz. EfficientNet_b0, ResNext50_32X4d, ConvNext_base using 6860 OPG collected from Faculty of Dentistry with meta data towards computational modelling of age threshold. We followed standard practices of Train/Test splits (85:15), judging performance of all models on the same independent Test data set of size 1029 OPGs. After Train/Test split, data augmentation with random horizontal flip, random rotation, and color jitters, etc were added on the fly to the Train
dataset. As only less than 20% of samples were belonging to Non-adult class, special set of hyper-parameters such as weighted loss, OPG reliant image shape (384X768), performance-based scheduler, etc. were incorporated while training of models. An ensemble of EfficientNet_b0 outperformed all other models with F1 score of 0.890 for Non-adult class, F1 score of 0.985 for Adult class, and F1 macro of 0.937 on an independent Test dataset of size 1029. We deployed the proposed model on a dynamic web server publicly available at: http://115.241.23.53:8000/ under tab “OPG Forensics” for radiologists to upload their OPG data and get results and Grad-CAM visualization within a few seconds. With our highly accurate model (F1 macro of 0.937), we aim to reduce the workload of forensic experts and support more efficient and objective judicial decision-making. This contribution marks an early and significant step toward AI-driven mathematical modeling of adult age threshold classification.

Biography: Dr. Suraiya Jabin is a Professor of Computer Science in Jamia Millia Islamia, central university in New Delhi, India. She has more than 23 years of experience in computer science teaching and research. Her research interests include Artificial Intelligence in Healthcare, Behavioural Biometrics, Smartphone signature biometrics, social media analysis, and computational biology. Her current work focuses on providing AI and deep learning-based solutions to problems in postgenomic biology. She has taught courses on machine learning, AI, digital image processing, deep learning, advanced DBMS, Compiler Design, Discrete Mathematics, etc. to Masters and UG, and Pre-PhD courses during her teaching career of 23 years.
She has contributed over 70 research articles in journals, conference proceedings, and book chapters, books, including 15 papers in SCIE/Scopus indexed journals, and 2 books on Machine Learning with Wiley India publisher. She owns an Indian patent titled “Mobile-Biometric Signature based Authentication System” dated Feb 2017, and 2 patents are under progress. She was PI for the departmental infrastructural grant Bioinformatics Infrastructure Facility Center funded by DBT, GoI for 8 years from 2012 to 2020. She successfully supervised 8 Ph.D. scholars, and currently, supervising 5 Ph.D. scholars working on various problems such as disease outbreak prediction using social media, crowd monitoring using deep learning, and Mental Task classification using EEG data, etc. She is a member of several professional bodies including ACM Professional Member since Feb 2024 (in recognition of her reviewer assignments for ACM journals), and a life member of Indian Society for Technical Education (ISTE) since 2005. She is an editorial board member of prestigious journals Nature Scientific Reports, Frontiers in Computer Science, IGI Global, etc. She is an active reviewer for various journals of IEEE, ACM, Springer, iScience, InderScience, Taylor & Francis, Sage, etc. Along with teaching & research, she has been serving several administrative responsibilities such as Teacher Placement Coordinator, UG/PG Curriculum Design Coordinator in the present, and non-resident warden, assistant proctor, etc. in the past in JMI.

 

Prof. Xiwen Zhang, Beijing Language and Culture University, China

Speech Title: Three views on intelligently extracting and generating information from image

Abstract: Due to pattern recognition and deep learning, various information can be extracted and generated from image. Our work has focused on using the proposed hierarchy models, local homogeneity, and adversarial generation.
Various digital images are processed, such as ones scanned from mechanical paper drawings and paper text, face images, portrait ones with line drawings, and microscopic bone marrow images.
Various information is extracted using the proposed hierarchy models. Graphics and their multi-levels compounded objects are extracted and recognized from images scanned from mechanical paper drawings using a hierarchy model of engineering drawings. Faces and their components are extracted from photos using a facial model.
Various information is extracted using the proposed local homogeneity. Karyocytes and their components from microscopic bone marrow images based on regional color features.
Various information is extracted and generated from image using cycle-Consistent adversarial networks. Text is separated from grid background using cycle-Consistent adversarial networks. Digital images of Chinese classical upper-class lady paintings are generated from images with line drawings using conditional generative adversarial networks.

Biography: Professor, Doctoral Supervisor, Beijing Language and Culture University Biography: XiWen Zhang is currently a full professor of Digital Media Department, School of Information Science, Beijing Language and Culture University.
Prof. Zhang worked as an associated professor from 2002 to 2007 at the Human-computer interaction Laboratory, Institute of Software, Chinese Academy of Sciences. From 2005 to 2006 he was a Post doctor advised by Prof. Michael R. Lyu in the Department of Computer Science and Engineering, the Chinese University of Hong Kong. From 2000 to 2002 he was a Post doctor advised by Prof. ShiJie Cai in the Computer Science and Technology department, Nanjing University. Prof. Zhang's research interests include pattern recognition, computer vision, and human-computer interaction, as well as their applications in digital image, video, and ink. Prof. Zhang has published over 60 refereed journal and conference papers. His SCI papers are published in Pattern Recognition, IEEE Transactions on Systems Man and Cybernetics B, Computer-Aided Design. He has published more than twenty EI papers.
Prof. Zhang received his B.E. in Chemical equipment and machinery from Fushun Petroleum Institute (became Liaoning Shihua University since 2002) in 1995, and his Ph.D. advised by Prof. ZongYing Ou in Mechanical manufacturing and automation from Dalian University of Technology in 2000.

 

Prof. Dinesh Goyal, Poornima Institute of Engineering & Technology, India

Speech Title: AI-Enabled Image Analysis Framework for Healthcare Applications

Abstract: Artificial intelligence (AI) has become one of the most effective medical image analysis tools, that allows it to interpret complicated visual patterns automatically to aid in clinical diagnosis and decision-making. The inconsistency of medical imaging data, preprocessing strategies, and fragmented evaluation practices, however, remain a limiting factor to the reliability and reproducibility of available methods. The paper describes a systematic AI-based image analysis system in healthcare, aimed at offering a framework of unified and repeatable medical image processing and analysis pipeline. The proposed architecture incorporates standardized image preprocessing, feature learning through deep learning, systematic model evaluation, and performance analysis in a modular architecture that can be tailored to be used in various imaging modalities. With the help of the available convolutional neural networks’ architecture, the framework is tested on representative healthcare imaging tasks, such as disease classification and medical image segmentation. The outcomes of the experiment indicate that deep learning models can be effectively utilized as diagnostic tools with regards to the accuracy, sensitivity, specificity, and AUC- ROC when integrated into a clear analytical framework. The results suggest that the suggested strategy enhances robustness and generalization on a wide range of imaging tasks and is still clinically relevant. The article highlights the significance of organized AI pipelines in the development of trusted and highly scalable medical image analysis to actual healthcare settings.

Biography: Dr. Dinesh Goyal, is an Professor in Computer Science & Engineering and is currently designated as Principal and Director at Poornima Institute of Engineering & Technology (PIET), Jaipur, with over 25 years of experience in teaching, research, and administration. He holds B.E, M.Tech, and Ph.D. degrees in Computer Science & Engineering. Dr. Goyal’s research interests include Cloud Security, Image Processing, Data Analytics, and Information Security, he has been pivotal in many turnkey projects & research and development activities. He has been pivotal in establishing advanced research labs—such as AICTE-sponsored IDEA Lab and Deep Learning Lab. He has been invited speaker and conference chair in various conferences organized globally in countries like China, Japan etc. His career includes successful organization of conferences, workshops, and faculty development programs, and he is an empanelled assessor for NAAC since 2021, contributing nationally to academic quality improvement. He has also received Grants-in-Aid for Research, Development, Conference & workshops, amounting more than Rs. 81 Lakh, from agencies like AICTE, TEQIP, ISTE etc, that include Combined Research Project, MODROBS, AICTE-IDEA Lab Dr. Goyal has published extensively and is known for driving innovation and outcome-based education in higher technical education. He has also completed his CMI level 5 Award in “Management and Leadership”, under AICTE-UKIERI Technical Leadership Development Program in association with Dudley University, United Kingdom. He has 36 Full patents published 2 Granted & 1 Copyright under his name. He has successfully published 16 edited books with big publishing giants like Springer, Wiley, IGI Global, Apple Academic Press, Taylor & Francis and Eureka. He has published 5 SCI & 116 Scopus and 52 Web of Science indexed papers & is editor of 2 SCI & 5 Scopus Indexed Journals, special issues. He has also attended more than 25 International Conferences & has been invited speaker for more than 15 Conferences & Seminars. He is Senior Member of IEEE, life member of ISC, CSI, IETE & ISTE and fellow member of ACM.

 

Assoc. Prof. Xiaolong Hu, Zhejiang University, China

Speech Title: LiDAR Imaging with High-Performance Fractal Superconducting Nanowire Single-Photon Detectors

Abstract: Traditional superconducting nanowire single-photon detectors (SNSPDs) with meander-nanowire structures can efficiently detect single photons in specific states of polarization while offering excellent timing performance and ultra-low dark count rates, making them widely used in quantum and classical faint-light detection. However, enabling SNSPDs to efficiently detect photons in arbitrary states of polarization while maintaining their other superior performance remains a challenge. To address this issue, we proposed the fractal SNSPDs capable of efficiently detecting incident photons in arbitrary states of polarization and developed: (1) a fractal SNSPD system coupled with single-mode fiber, featuring plug-and-play operation, achieving a system detection efficiency (SDE) of 91% for arbitrary polarization states at a wavelength of 1540 nm; (2) an 8-channel fractal SNSPD system operating in the 930–940 nm band, with an average SDE of 90% for arbitrary polarization states across the eight channels and a maximum SDE of 96% in the channel with the highest detection efficiency; (3) a fractal SNSPD with a large photosensitive area, coupled with a 50-μm core-diameter multimode fiber, achieving an SDE of 78% for arbitrary polarization states at a wavelength of 1530 nm. Using fractal SNSPDs, we demonstrated the following applications: (1) photon time-of-flight (ToF) LiDAR imaging, (2) full-Stokes LiDAR imaging, (3) non-line-of-sight imaging, (4) high-precision dual-comb ranging, and (5) three-dimensional single-pixel imaging.

Biography: Dr. Xiaolong Hu is a tenured associate professor at the College of Information and Electronic Engineering, Zhejiang University. His research focuses on micro- and nano- optoelectronic devices and quantum photonic devices, and he has achieved a series of research results in the field of SNSPDs: (1) He proposed waveguide-integrated SNSPDs, which has now become the mainstream device structure for SNSPDs on integrated quantum optical chips; (2) He revealed two mechanisms of device timing jitter of SNSPDs and developed the device physics of SNSPDs; (3) He proposed and developed fractal SNSPDs with high system detection efficiency and low timing jitter for incident photons in all states of polarization, and developed practical systems that are fiber-coupled and plug-and-play; (4) He applied the fractal SNSPDs to dual-comb precision ranging, full-Stokes lidar imaging, infrared non-line-of-sight imaging, and three-dimensional single-pixel imaging. The fractal SNSPD systems have been used in many institutions such as Peking University, Tsinghua University, Zhejiang University, Sun Yat-sen University, Wuhan University of Technology, and Songshan Lake Materials Laboratory. The research results have been published in journals such as Nature, Nature Photonics, and Nature Nanotechnology. The fractal SNSPDs have received medium coverage from Nature Photonics (News & Views), the Optica, IEEE Photonics Society, IEEE Spectrum, and Science and Technology Daily; the fractal SNSPDs have won the golden medal at the 28th National Invention Exhibition, the first-grade prize of the Invention and Entrepreneurship Award of the China Association of Inventions, the gold medal at the 50th Geneva International Invention Exhibition, and a nomination for the Chip 10 Science Award in 2024. Dr. Hu is now serving as an associate editor of Optics Express.

 

Senior Lecturer Anwaar Ulhaq, Central Queensland University, Australia

Speech Title: From Artificial to Real Intelligence

Abstract: Artificial intelligence can recognise patterns and learn from data, but it still lacks the flexibility and efficiency seen in biological systems. This talk explores how real intelligence can be understood by studying living neural structures. I will present recent work on brain organoids, neural imaging, and computational modelling, including our research on organoid mitosis analysis using computer vision. These studies show how biological systems organise, adapt, and learn in ways that current algorithms do not. By combining ideas from neuroscience, imaging, and machine learning, we can design models that are more robust and closer to real intelligence. The talk also outlines future research directions connecting computer vision, brain science, and neuromorphic computing.

Biography: Dr Anwaar Ulhaq is a Senior Lecturer in Artificial Intelligence at Central Queensland University's Sydney Campus, Australia. He holds a PhD in Artificial Intelligence from Monash University, Australia; a Graduate Certificate in Machine Learning and Artificial Intelligence from Massachusetts Institute of Technology, USA; and a professional certificate in Business Analytics from Harvard Business School. Additionally, he completed the Oxford Executive Leadership Program at the University of Oxford's Said Business School. He is the current President of the Australian Pattern Recognition Society and is a member of the Australian Academy of Sciences, the Australian Computer Society, and IEEE Signal Processing. With a diverse academic background, Dr Ulhaq has contributed significantly to the field of computer vision, serving as the General Chair of DICTA 2023, a major Australian computer vision conference. He is an associate editor of IEEE Transactions on Image Processing (TIP). He has published over 100 peer-reviewed journal and conference papers, received teaching and research excellence awards, and secured approximately $3.15 million in research funding as a principal investigator and co-investigator. Dr Ulhaq's research interests span computer vision, image and signal processing, responsible AI, and remote sensing, and his work has been featured in national and international media more than 30 times, underscoring the relevance and significance of his contributions.

 

Assoc. Prof. Ts. Dr. Mas Rina Mustaffa, Universiti Putra Malaysia, Malaysia

Speech Title: From Fabric Patterns to Garment Visualisation Using Image Processing and Generative Modelling

Abstract: A considerable number of individuals rely on visual representation when making design-related decisions, creating a need for tools that can effectively preview how selected fabrics will appear when formed into garments. This study proposes a computational framework grounded in image processing and generative modelling to simulate the transformation of fabric patterns into traditional attire, with a focus on structured garment visualisation. The approach utilises binary silhouette mapping combined with morphological image processing techniques, including erosion and dilation, to generate refined two-dimensional garment representations. To enhance realism and allow flexible customisation, a diffusion-based inpainting model is incorporated to produce visually coherent and context-aware fabric patterns. The framework is implemented within a mobile-based prototype, enabling users to interactively apply and adjust fabric designs in near real time. Evaluation results indicate that the system performs effectively in terms of usability, functional capability, and user satisfaction. Overall, the proposed framework demonstrates the potential of integrating image processing techniques with generative modelling for practical garment visualisation, supporting more informed decision-making and improving the overall fabric selection experience.

Biography: Assoc. Prof. Ts. Dr. Mas Rina Mustaffa is an Associate Professor at the Multimedia Department, Faculty of Computer Science and Information Technology, Universiti Putra Malaysia (UPM), and a registered Professional Technologist (Ts.). She received her PhD in Multimedia Systems from Universiti Putra Malaysia in 2012. A Senior Member of IEEE, she specializes in computer vision, multimedia analytics, pattern recognition, and multimedia information retrieval, with a focus on AI-driven solutions for education, agriculture, and intelligent systems.
She has led and collaborated on major national and international projects, including Malaysia’s FRGS grants and the European Union’s Horizon 2020 ULTRACEPT initiative, and recently completed a research secondment at the University of Leicester, UK (2024). Dr. Mas Rina has published extensively in high-impact journals and international conferences (IEEE, ACM, Springer) and is a recipient of multiple Best Paper and Presentation Awards. She is Editor-in-Chief of the Journal of Intelligent Media Computing, Vice President of PECAMP Malaysia (2025–2027), and currently serves as Publication Chair and Track Chair for several international conferences, advancing global research in intelligent multimedia representation and retrieval.

 

Assoc. Prof. Syed Farooq Ali, University of Management & Technology, Pakistan

Speech Title: A boosting framework for human posture recognition using spatio-temporal features along with radon transform

Abstract: Automatic human posture recognition in surveillance videos has real world applications in monitoring old-homes, restoration centers, hospitals, disability, and child-care centers. It also has applications in other areas such as security and surveillance, sports, and abnormal activity recognition. Human posture recognition is a challenging problem due to occlusion, background clutter, illumination variations, camouflage, and noise in the captured video signal. In the current study, which is an extension of our previous work (Ali et al. Sensors, 18(6):1918, 2018), we propose a novel combination of a number of spatio-temporal features computed over human blobs in a temporal window. These features include aspect ratios, shape descriptors, geometric centroids, ellipse axes ratio, silhouette angles, and silhouette speed. In addition to these features, we also exploit the radon transform to get better shape based analysis. In order to obtain improved posture classification accuracy, we used J48 classifier under a boosting framework by employing the AdaBoost algorithm. The proposed algorithm is compared with eighteen existing state-of-the-art approaches on four publicly available datasets including MCF, UR Fall detection, KARD, and NUCLA. Our results demonstrate the excellent performance of the proposed algorithm compared to these existing methods.

Biography: Dr. Syed Farooq Ali has around 18+ years of teaching and research experience. His areas of specialization include Computer Vision, Digital Image Processing, Medical Imaging, and Image & Video Coding. He did his BS from NUCES- FAST, Lahore, and later earned an MS degree from LUMS with Dean’s Honor List and 4th position out of a batch of around 130 graduate students. During his stay in MS, he was on a LUMS fellowship. He also completed his MS from Ohio State University, (Col.), USA. Moreover, he passed the Ph.D. Comprehensive Exam (Qualifier Exam) from Ohio State University, Col. USA. Later, he transferred his Ph.D. from Ohio State University USA to UMT and completed it.
During his graduate studies at LUMS and Ohio State University, he studied a total of 34 courses. Out of these, he studied ten courses related to his depth area (area of interest) that included Computer Vision, Medical Imaging, Image and Video Coding, Digital Image Processing, Artificial Intelligence, Pattern Recognition, Machine Learning, Computer Graphics, Signals and Systems, and Data Mining.
Currently, he is Associate Professor and Chair Vision and Image Processing Research Group at UMT. He also served two times as a Director Projects. Dr. Ali has around 39 publications including 23 journals, a book chapter, and 15 international conference papers. Dr. Ali is also a reviewer of many international conferences and journals. He also won 8 Fundings from National Grassroots ICT Research Initiative (NGIRI) for Final Year Projects.

 

 

Senior Lecturer Philippe Durand, Conservatoire National des Arts et Métiers, France

Speech Title: Morphological and Topological Methods for Urban Extraction from Noisy Radar Imagery: Application to the City of Le Luc (France)

Abstract: This invited lecture presents a complete mathematical framework for extracting urban structures—built-up areas, residential blocks, and transport networks— from highly noisy radar imagery acquired over the city of Le Luc in South-Eastern France. Radar images of semi-urban environments are complex to interpret due to speckle noise, non-Gaussian backscatter distributions, and the multiple scattering mechanisms generated by buildings and roads. Our goal is to derive a coherent representation of the urban fabric comparable to the structures visible in the optical aerial photograph, while relying exclusively on the raw radar acquisition.
We develop an integrated approach combining two mathematical paradigms: (1) Mathematical Morphology, including Alternating Sequential Filters, granulometry, directional openings, and morphological skeletonization; and (2) Topological Data Analysis (TDA), based on persistent homology, H0–H1 generators, and Wasserstein distances. Morphology provides a multi-scale geometric description, capable of reducing speckle and isolating coherent radar responses associated with buildings and road segments. TDA offers complementary information on the global structure of urban textures, revealing stable topological patterns (loops, cavities, block organisation) which remain robust under noise and local fluctuations.
Applied to the Le Luc dataset, the combined methodology enables the recovery of major urban zones (historical nuclei VN1, VN2, VN3), pavillonnaire housing areas, dense HLM blocks, and the main transport axes (highway, major roads, railway). The morphological segmentation is strongly supported by the TDA signatures: urban regions exhibit a rich distribution of persistent H1 generators, while background areas show simple connectivity profiles dominated by H0 components.
This talk highlights the synergy between geometry and topology for radar image analysis, demonstrating that the fusion of morphological filtering with TDA produces a robust, interpretable, and noise-tolerant pipeline for urban structure extraction, even under severe speckle conditions. Perspectives include extensions to multi-temporal radar sequences, polarimetric acquisitions, and integration into near-real-time environmental monitoring systems.

Biography: Philippe Durand is Senior Lecturer in the Mathematics and Statistics Department of the National Conservatory of Arts and Crafts in the Mathematical and Numerical Modeling Department (M2N), he works on the interaction between mathematical engineering and the theoretical tools of mathematics including usage has been increasing since the introduction of modern mathematics in the early sixties. He is interested in the mathematization of gauge theories in physics and string theory, he also works on tensor analysis applied to networks as well as the application of topological and statistical methods to image processing. In image processing, he used remote sensing images and especially radar images, he invested different methods of pattern recognition, and in particular the tools of mathematical morphology for the extraction of texture information. Currently I am focusing on the use of topological data analysis and different approaches to applying classical or quantum neural networks to image processing. He published his results in various journals of math-ematical engineering, and various proceedings of image processing conferences. Philippe Durand is assistant professor in Department of Mathematics (mod´elisation math´ematique et num´erique), Conservatoire National des Arts et M´etiers, 292 rue Saint Martin, 75141 Paris, FRANCE, (e-mail: philippe.durand@lecnam.net).

 

Asst. Prof. Zhennong Chen, Xi'an Jiaotong-Liverpool University, China

Speech Title: AI-empowered brainCT motion correction

Abstract: Motion artifacts remain one of the primary sources of image degradation in computed tomography (CT). Patient movement violates the static-object assumption underlying conventional reconstruction algorithms, leading to image blurring, distortion, and reduced diagnostic accuracy. To address this challenge, we have developed two complementary AI-based approaches that operate in the sinogram and image domains, respectively. In the sinogram domain, we designed a convolutional neural network (CNN) that takes partial-angle reconstructions (PARs) as input to estimate patient motion and perform motion compensation prior to final image reconstruction. In the image domain, we proposed a diffusion generative model incorporating histogram equalization and an elucidated diffusion framework for motion artifact correction. Furthermore, we conducted a human observer study on a real-world portable brain CT scan dataset to validate that the diffusion-based correction not only enhances image quality but also preserves lesion detectability.

Biography: Dr. Zhennong Chen is an Assistant Professor at the School of Artificial Intelligence and Advanced Computing (AIAC), Xi’an Jiaotong-Liverpool University. He received both his Ph.D. and Bachelor’s degree in Bioengineering from the University of California, San Diego, and subsequently worked as a Postdoctoral Research Fellow at Harvard Medical School and Massachusetts General Hospital. Dr. Chen’s research focuses on developing innovative AI methods for medical imaging, with particular emphasis on computed tomography (CT) and cardiovascular imaging. He has received the Trainee Research Prize at RSNA and the Best All-Conference Paper Finalist at SPIE Conference.

 

Asst. Prof. Zhi Lu, Tsinghua University, China

Speech Title: Leveraging spatial-angular redundancy for self-supervised denoising of 3D fluorescence imaging without temporal dependency

Abstract: Photon noise is a major bottleneck for extracting reliable biological information from fluorescence microscopy, especially when imaging fast and volumetric biological dynamics. However, most existing self-supervised denoising strategies depend on repeated measurements in time or strong spatial assumptions, which inevitably reduce temporal or spatial fidelity. Here we introduce LF-denoising, a transformer-based framework that exploits the rich spatial–angular redundancy encoded in light field measurements, enabling accurate denoising without requiring temporal repetition. Across simulations and diverse intravital experiments, LF-denoising robustly improves 3D imaging quality under extremely low excitation power, and generalizes across species including zebrafish, Drosophila and mice. By preserving the true temporal evolution of biological processes while substantially reducing noise, LF-denoising opens a path toward more reliable and accessible high-speed 3D imaging for quantitative biology, with broad applications.

Biography: Dr. Zhi Lu is currently an Assistant Professor at Tsinghua University. He received the B.S. and Ph.D. degrees in Control Science and Engineering from Tsinghua University, Beijing, China, in 2018 and 2023, followed by postdoctoral research from 2023 to 2025. His research interests include computational imaging and intelligent microscopy. In recent years, He has published papers in journals including Cell, Nature Methods, Nature Biotechnology, Nature Protocols and Nature Communications, with over twenty granted patents. He is honorably on Forbes China 30 Under 30 List (2023), and supported the National Postdoctoral Program for Innovative Talents (2023) and by the Young Top-notch Talent of National High-Level Talent Special Support Program (2024). Additionally, he received SAIL awards at the World Artificial Intelligence Conference (2022, 2024), the Dimitris N. Chorafas Prize (2024), Best Paper Award in Computational Optics at IBCS (2023), and the Award in Ten Advances for Optics in China (2022).

Dr. Loc Nguyen, Sunflower Soft Company, Vietnam

Speech Title: Is matrix neural network the alternative of convolutional neural network?

Abstract: Currently, deep learning is the most important and popular methodology in artificial intelligence (AI) and artificial neural network (ANN) is the foundation of deep learning. The main drawback of ANN is the boom problem of a huge number of parametric weights when ANN in deep learning establishes a large number of hidden layers. The excellent solution for image processing within context of deep learning is convolutional neural network (CNN) equipped filtering kernel. Another solution of the boom problem is that large parametric weight vector is organized as matrix, which leads to a so-called matrix neural network (MNN). Computation cost of MNN is decreased significantly in comparison with ANN but it is necessary to test the main hypothesis “whether MNN is the alternative of CNN”. Moreover, transformer which is the new trend in AI and deep learning, which aims to improve/replace traditional ANN by self-supervised learning, in which attention is the significant mechanism of self-supervised learning. Therefore, the implicit deep meanings of attention and filtering kernel are similar, which represents feature of data, which does not go beyond parametric weights too. In general, the research has two goals: 1) explaining and implementing ANN, CNN, and transformer (attention) and 2) applying analysis of variance (ANOVA) into evaluating the effectiveness of ANN, CNN, and transformer (attention) within context of image classification. The ultimate result is that it is not asserted that MNN is the alternative of CNN but MNN can be an optional choice for implementing ANN instead of focusing on the unique CNN solution. Moreover, the incorporation of MNN and attention in implementing transformer produces a compromising solution of high performance and computational cost.

Biography: Loc Nguyen is an independent scholar from 2017. He holds Master degree in Computer Science from University of Science, Vietnam in 2005. He holds PhD degree in Computer Science and Education at Ho Chi Minh University of Science in 2009. His PhD dissertation was honored by World Engineering Education Forum (WEEF) and awarded by Standard Scientific Research and Essays as excellent PhD dissertation in 2014. He holds Postdoctoral degree in Computer Science from 2013, certified by Institute for Systems and Technologies of Information, Control and Communication (INSTICC) by 2015. Now he is interested in poetry, computer science, statistics, mathematics, education, and medicine. He serves as reviewer, editor, speaker, and lecturer in a wide range of international journals and conferences from 2014. He is volunteer of Statistics Without Borders from 2015. He was granted as Mathematician by London Mathematical Society for Postdoctoral research in Mathematics from 2016. He is awarded as Professor by Scientific Advances and Science Publishing Group from 2016. He was awarded Doctorate of Statistical Medicine by Ho Chi Minh City Society for Reproductive Medicine (HOSREM) from 2016. He was awarded and glorified as contributive scientist by International Cross-cultural Exchange and Professional Development-Thailand (ICEPD-Thailand) from 2021 and by Eudoxia Research University USA (ERU) and Eudoxia Research Centre India (ERC) from 2022. He has published 101 papers and preprints in journals, books, conference proceedings, and preprint services. He is author of 5 scientific books. He is author and creator of 10 scientific and technological products.

 

 

 

 

 

INVITED SPEAKERS of AAIP2025

Philippe Durand	Yanglong Lu	Sergii Khlamov	Abhishek Shukla
Conservatoire National des Arts et Métiers, France	The Hong Kong University of Science and Technology, Hong Kong, China	Kharkiv National University of Radio Electornics, Ukraine	Syracuse University, NY, USA