Jamalia Sultana
Jamalia Sultana
CS PhD (3rd year)|Researcher
I am a Graduate Student (3rd year in PhD) in the Department of Computer Science (CS) at Stony Brook University. My research interests broadly lie in Multimodal AI, Foundation Models, Representation Learning, VLM, and Medical Image Analysis.
Ph.D. in Computer Science at Stony Brook University (August 2023- Current)
M.Sc.Engg. in Computer Science and Engineering (CSE) at Bangladesh University of Engineering and Technology (BUET) (July 2021- June 2023)
B.Sc. in Computer Science and Engineering (CSE) at Bangladesh University of Engineering and Technology (BUET) (February 2016 - March 2021)
Viqarunnisa Noon School and College (HSC'15, SSC'13)
Multimodal AI.
Foundation Models.
Representation Learning.
Vision-Language Models.
Segmentation.
Scanpath Prediction.
Tracking.
Medical Image Analysis.
A multimodal GNN that uses chest X-rays, eye gaze, and reports to improve disease localization and classification, paired with gaze-aware prompts and a fine-tuned LLM to write region-specific texts along with an interactive Q&A with attention maps to help trainees read images.
Building a large-scale multimodal landfill dataset combining egocentric video, synchronized speech, and interaction cues, with curated taxonomies and time-aligned annotations for waste sorting, enabling research on segmentation in cluttered scenes and foundation models.
Automatic Solid Waste Annotation and Segmentation in Landfill
Building a large-scale multimodal landfill dataset combining egocentric video, synchronized speech, and interaction cues, with curated taxonomies and time-aligned annotations for waste sorting, enabling research on segmentation in cluttered scenes and foundation models.
Developing a human-in-the-loop pipeline around promptable segmentation foundation models (SAM2) to generate masks from weak supervision (speech cues and sparse prompts) with minimal manual labeling.
Extending from frame-level masks to video object segmentation/object tracking via mask propagation across adjacent frames to stabilize labels and reduce re-prompting cost.
Improving robustness in clutter/occlusion using open-vocabulary retrieval (e.g., CLIP prototypes) to auto-suggest prompts and pseudo-masks, then iteratively refining with human corrections.
Scanpath Prediction for Thoracic Diseases
Built a transformer-based model to predict radiologist-like scanpaths for thoracic diseases.
Quantified and analyzed differences between human expert and AI-generated scanpaths to improve interpretability.
Evaluated the impact of simulated scanpaths on downstream segmentation/classification tasks, demonstrating measurable performance gains.
2026
Graduate Research Assistant, Dept. of Computer Science, Stony Brook University
Supervisor: Prof. Dr. Zhaozheng Yin
Developed an SSL that outperforms foundation-model baselines using <1% of training data by distilling radiologists’ attention maps to learn expert visual priors for thoracic disease identification; demonstrated gains on downstream localization/segmentation tasks. Introduced a Teacher's Assistant in the process to bridge the knowledge distillation from multiple models to one singular model. [in submission ECCV'26]
Built a gaze-guided training assistant for medical imaging: a multimodal GNN that uses chest X-rays, eye gaze, and reports to improve disease localization and classification, paired with gaze-aware prompts and a fine-tuned LLM to write region-specific texts along with an interactive Q&A with attention maps to help trainees read images. [ISBI '26, Oral Presentation]
Developed a novel graph-based neural network to predict diseases from medical images by integrating the expert's eye gaze pattern and the expert's speech transcriptions. [ACCV '24] (NSF grants: ECCS-2026357 and ECCS-2025929)]
2023
Graduate Research Fellow, Bangladesh University of Engineering and Technology (BUET), Computer Science and Engineering (CSE) - July 2021 - Jun 2023. Only student from CSE to receive this honor.
Automated Framework for Bone Isolation and Reconstruction in Bone Morphology Study. Supervisor: Prof. Dr. Mahmuda Naznin.
Jamalia Sultana, Dr. A. B. M. Alim Al Islam, Analyzing Reliability of Medical Anomaly detecting Deep Learning Models for Medical Data Corruption by Humans.
Md. Shaifur Rahman, Jamalia Sultana, Mohammed Nahiyan Uddin, Rifat Shahriyar, Sadia Sharmin, Mahmuda Naznin, Leveraging Operations of A Mobile Agent in A Smart Sensing System through Efficient Path Planning.
2021
Research Assistant (Part-Time), United International Univerisity (UIU), Computer Science and Engineering (CSE), at Diabetes Project - August. 2021 - June 2022.
Blood Glucose Prediction from Daily Food Intake and Activities. Grant No.- UIU/IAR/01/2021/SE/11, PI-Md. Benzir Ahmed (CSE, UIU), Co-PI- Mahmuda Naznin (CSE, BUET), Md. Eunus Ali (CSE, BUET), Mehedy Masud (UAE University), (2021-2023) . Supervisors: Prof. Dr. Mahmuda Naznin (Co-PI), Benzir M. Ahmed (PI), Prof. Dr. Muhammed Eunus Ali (Co-PI), Prof. Dr. Mehedy Masud (Co-PI).
Jamalia Sultana, Benzir M. Ahmed, Mohammad Mehedy Masud, Dr. Mahmuda Naznin, Mohammed Eunus Ali, A Novel Deep Learning CNN-based Framework for Calorie Estimation from Food Images.
2020
Undergraduate Research Student, Bangladesh University of Engineering and Technology (BUET), Computer Science and Engineering (CSE) - January 2020 - May 2021.
Speech Emotion Recognition from Voice Data. Supervisor: Prof. Dr. Mahmuda Naznin.
2026
Jamalia Sultana, Ruwen Qin, Zhaozheng Yin, RadGaze-LLM: Anatomical Region-Grounded Radiology Report Generation via Learning from Expert Gaze, International Symposium on Biomedical Imaging (ISBI), 2026. [Oral Presentation]
2024
Jamalia Sultana, Ruwen Qin, Zhaozheng Yin, Seeing Through Expert’s Eyes: Leveraging Radiologist Eye Gaze and Speech Report with Graph Neural Networks for Chest X-ray Image Classification, 17th Asian Conference on Computer Vision (ACCV), 2024.
2023
Jamalia Sultana, Dr. Mahmuda Naznin, Dr. Tanvir Faisal, Semi-Supervised Learning Based Femur Segmentation from QCT Images, 22nd International Conference on Machine Learning and Application (IEEE ICMLA), 2023.
Jamalia Sultana, Dr. Mahmuda Naznin, Dr. Tanvir Faisal, SSDL–An efficient automated semi-supervised deep learning approach for patient-specific 3D reconstruction of proximal femur from QCT images, Medical & Biological Engineering & Computing (MBEC), 2023.
Jamalia Sultana, Benzir MD. Ahmed, Mohammad Mehedy Masud, A. K. Obidul Huq, Mohammed Eunus Ali, and Mahmuda Naznin, A Study on Food Value Estimation From Images: Taxonomies, Datasets, and Techniques, IEEE Access, 2023.
Jamalia Sultana, Dr. Mahmuda Naznin, Dr. Tanvir Faisal, Automated End-to-end Segmentation of Proximal Femur, in 18th International Symposium on Computer Methods in Biomechanics and Biomedical Engineering (CMBBE), 2023. Accepted for Poster Presentation.
Jamalia Sultana, Dr. Mahmuda Naznin, Dr. Tanvir Faisal, Automated Femur Isolation In CT Scan Images By Generating Binary Mask-a Critical Step For 3d Femur Reconstruction, in Orthopedic Research Society (ORS), 2023. Accepted for Poster Presentation.
2022
Jamalia Sultana, Dr. A. B. M. Alim Al Islam, Analyzing Fault-tolerance in Deep Learning Models for Medical Data Corruption by Humans, in 9th International Conference on Networking, Systems and Security (NSysS), 2022. Received the Champion Student Poster Award for Student Poster Presentation in NSysS 2022.
Jamalia Sultana, Dr. Mahmuda Naznin, Breaking the Barrier with a Multi-Domain SER, in IEEE 46th Annual Computers, Software, and Applications Conference, (IEEE COMPSAC), 2022.
2020
Jamalia Sultana, Dr. Mahmuda Naznin, Impact of Biological Gender on Emotion Detection in Voice Data in Grace Hopper Conference (GHC), 2020. Published as Poster Session. Selected for ACM Student Research Competition.
Guest Presenter
BMI 514: Imaging Informatics Analysis, Stony Brook University
Teaching Assistant
CSE 334: Intro. to Multimedia
CSE 310: Computer Networks
Reviewer
CVPR, ICCV, NeurIPS, CVMI, ACCV
Received Champion Student Poster Award for Student Poster Presentation in NSysS 2022.
Selected as Student Volunteer at ACM UbiComp/ISWC 2022.
Graduate Research Fellowship from BUET 2021.
Training in CITI Program, Western Washington University, 2021.
Selected for ACM Student Research Competition, Grace Hopper Conference (GHC), 2020.
Selected for Study Tour at NASA, Kennedy Space Center in Florida, USA (Summer Camp for International Students, 2014).
Merit Scholarship in Higher Secondary Certificate Examination, Dhaka Education Board, 2015, 2013.
Data Science: Dr. Steven Skiena
Meta Heuristics: Dr. M. Sohel Rahman
Network Science: Dr. Md. Saidur Rahman
Bioinformatics Algorithms: Dr. Md Shamsuzzoha Bayzid
Advanced Dependable and Fault-Tolerant Computer Systems: Dr. A. B. M. Alim Al Islam Islam
Pattern Recognition
Artificial Intelligence
Data Structures and Algorithms
Fault Tolerant Systems
Computer Network and Security
Operating System