Research & Publication

My current work asks how medical vision-language systems represent clinical information, perform on rare findings, and support trustworthy interpretation and evaluation.

Browse publications
01 · Medical Vision-Language Models

Understanding multimodal models in radiology

I study how multimodal foundation models represent radiology images and clinical concepts. CTVLM-Inspect evaluates chest-CT vision-language models across classification, calibration, retrieval, reconstruction consistency, and interpretability.

Current questions

  • What clinical information do chest-CT vision-language models represent reliably?
  • How do classification, retrieval, calibration, and reconstruction-consistency evaluations complement one another?
  • Where do model behaviour and clinical concepts diverge across tasks and datasets?
CTVLM-Inspect workflow from chest CT model inputs to a multi-metric evaluation suite
CTVLM-Inspect evaluation workflow.
02 · Long-Tailed Medical AI

Making rare findings visible in evaluation

I study rare-finding recognition, cross-dataset behaviour, calibration, grounding, and adaptation strategies for chest X-ray and CT models.

Current questions

  • How does model performance change from common findings to rare findings?
  • Which evaluations expose underrepresented-class failures that aggregate metrics conceal?
  • How do adaptation, grounding, and calibration behave across datasets with different finding frequencies?
Map of long-tailed radiology applications, research problems, and methods
Long-tailed chest-radiology research map.
03 · Interpretable & Reliable AI

Interpreting model behaviour under shift

I study representation and explanation behaviour under class imbalance and distribution shift, alongside model robustness, grounding, and calibration in radiology.

Current questions

  • What do interpretation methods reveal about biomedical vision-language models under imbalance?
  • How stable are explanations and representations when data shift between settings?
  • How can evaluation distinguish plausible explanations from reliable model behaviour?
BiomedCLIP data-processing, model configuration, and evaluation diagram
BiomedCLIP study workflow.

Publications

Browse by year or focus on a research area.

Download BibTeX ↓
Showing 27 of 27

2026

DepthPulse+: A Depth and Vital Sign Based Method for Face Presentation Attack Detection

Nafiz Sadman, Furkan Alaca, Farhana Zulkernine

ICC 2026-IEEE International Conference on Communications · pp. 1–6

CVCybersecurity
Read summary
Combines depth maps and remote photoplethysmography (rPPG) signals to detect face presentation attacks such as print, replay, and mask attacks on commodity cameras. Improved accuracy under cross-device evaluation.

PLeDO: Pain level detection for osteoarthritis from EMR data

Yuhao Chen, Jiahao Cai, Nafiz Sadman, Farhana Zulkernine, John Queenan, David Barber

Intelligent Data Analysis · pp. 1088467X261472133

Healthcare

2025

Visualizing temporal interval hierarchies

Nafiz Sadman, Nastaran Kianersi, Sean Kauffman

NASA Formal Methods Symposium · pp. 333–340

Visualization
Read summary
A visualization research tool for streaming PID-controller data using Python Socket.IO and D3 to expose temporal interval hierarchies for live formal-methods analysis.

Interpreting Biomedical VLMs on High-Imbalance Out-of-Distributions: An Insight into BiomedCLIP on Radiology

Nafiz Sadman, Farhana Zulkernine, Benjamin Kwan

arXiv preprint arXiv:2506.14136

CVMultimodalHealthcare
Read summary
Probes BiomedCLIP behaviour on imbalanced out-of-distribution radiology data and uses GradCAM-based interpretation to examine failure modes and clinical robustness.

Diagnosis and Management of Alzheimer’s Disease in Primary Care: A Real-World Study in Ontario, Canada

Zahinoor Ismail, Melanie Wilson, Hany Khalifa, Daniela Belovich, Eileen Shaw, Tram Pham, Suzanne McMullen, Yuhao Chen, Nafiz Sadman, Jackson Cai, et al.

Journal of Primary Care & Community Health · vol. 16 · pp. 21501319251363156

Healthcare
Read summary
A real-world Ontario EHR study of Alzheimer's diagnosis and management pathways. Nafiz contributed EHR extraction, k-means clustering, and logistic regression.

SiamesePhish: An Advanced AI-powered URL Embedding Approach with Synthetic URLs for Phishing Detection

Austin Carthy, Nafiz Sadman, Farhana Zulkernine

International Conference on Risks and Security of Internet and Systems · pp. 365–375

NLPCybersecurity

2024

Software engineering challenges in the area of testing and project management

Azaz Ahamed, Nafiz Sadman, Touseef Aziz Khan, Mahfuz Ibne Hannan

Thesis, IUB, CSE

Software Engineering

2023

Behavioral recommendation engine driven by only non-identifiable user data

Kishor Datta Gupta, Nafiz Sadman, Akib Sadmanee, Md Kamruzzaman Sarker, Roy George

Machine Learning with Applications · vol. 11 · pp. 100442

Recommender Systems

The rise of remote project management-a new norm?-a survey on it organizations in bangladesh

Azaz Ahamed, Touseef Aziz Khan, Nafiz Sadman, Mahfuz Ibne Hannan, Nujhat Nahar, Mahady Hasan

Independent University, Bangladesh

Software Engineering

Vulnerability of open-source face recognition systems to blackbox attacks: A case study with insightface

Nafiz Sadman, Kazi Amit Hasan, Elyas Rashno, Furkan Alaca, Yuan Tian, Farhana Zulkernine

2023 IEEE Symposium Series on Computational Intelligence (SSCI) · pp. 1164–1169

CVCybersecurity
Read summary
A black-box attack case study on InsightFace examining transferability of perturbations across open-source face-recognition systems.

2022

A Safer Approach to build Recommendation Systems on Unidentifiable Data

Kishor Datta Gupta, Akib Sadmanee, Nafiz Sadman

Proceedings of the 14th International Conference on Agents and Artificial Intelligence · vol. 3 · pp. 589–596

Recommender Systems

Interpretable learning model for lower dimensional feature space: a case study with brown spot detection in rice leaf

Md Amit Khan, Nafiz Sadman, Kishor Datta Gupta, Jesan Ahammed Ovi

2022 IEEE 12th Annual Computing and Communication Workshop and Conference (CCWC) · pp. 0428–0434

CV

Promise of AI in DeFi, a systematic review

Nafiz Sadman, Md Manjurul Ahsan, Abdur Rahman, Zahed Siddique, Kishor Datta Gupta

Digital · vol. 2 · pp. 88–103

Finance

HeteroGenius: An Improvised ‘Intelligence’in Heterogeneous Graph Transformers

Nafiz Sadman, Akib Sadmanee, Kishor Datta Gupta, Roy George

2022 21st IEEE International Conference on Machine Learning and Applications (ICMLA) · pp. 849–854

Machine Learning

Automated testing: Testing top 10 owasp vulnerabilities of government web applications in bangladesh

Azaz Ahamed, Nafiz Sadman, Touseef Aziz Khan, Mahfuz Ibne Hannan, Farzana Sadia, Mahady Hasan

ICSEA 2022 · vol. 56

CybersecuritySoftware Engineering

2021

Medical Speciality Detector Application (MSDA) Using AI

Nafiz Sadman, Md Habibur Rahman, Md Ariful Haque, Kishor Datta Gupta

Association for the Advancement of Artificial Intelligence, Menlo Park, CA, USA

NLPHealthcare

Understanding the pandemic through mining covid news using natural language processing

Nafiz Sadman, Nishat Anjum, Kishor Datta Gupta, MA Parvez Mahmud

2021 IEEE 11th Annual Computing and Communication Workshop and Conference (CCWC) · pp. 0362–0367

NLPHealthcare

Recommend speciality doctor from health transcription: Ensemble machine learning approach

Nafiz Sadman, Md Habibur Rahman, Sumaiya Tasneem, Mohd Ariful Haque, Kishor Datta Gupta

2021 IEEE 11th Annual Computing and Communication Workshop and Conference (CCWC) · pp. 0967–0972

NLPHealthcareRecommender Systems

A review on comparative remarks, performance evaluation and improvement strategies of quadrotor controllers

Rupal Roy, Maidul Islam, Nafiz Sadman, MA Parvez Mahmud, Kishor Datta Gupta, Md Manjurul Ahsan

Technologies · vol. 9 · pp. 37

Robotics

Leveraging Reinforcement Learning to build a Recommendation System for Incognito mode Users

Kishor Datta Gupta, Nafiz Sadman

ICML 2021 Workshop on Representation Learning for Finance and e-Commerce Applications

Recommender Systems

Promise of AI in Defi, A Literary Analysis

Nafiz Sadman, Abdur Rahman, Kishor Datta Gupta

Preprints

Finance

Utilizing computational complexity to protect cryptocurrency against quantum threats: a review

Kishor Datta Gupta, Abhijit Kumar Nag, Md Lutfar Rahman, MA Parvez Mahmud, Nafiz Sadman

IT Professional · vol. 23 · pp. 50–55

CybersecurityFinance

2020

ADCR: An Adaptive TOOL to select” Appropriate Developer for Code Review” based on Code Context

Nafiz Sadman, Md Manjurul Ahsan, MA Parvez Mahmud

2020 11th IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON) · pp. 0583–0591

Software Engineering

Stylometry as a reliable method for fallback authentication

Nafiz Sadman, Kishor Datta Gupta, Md Ariful Haque, Sajib Sen, Subash Poudyal

2020 17th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON) · pp. 660–664

NLPCybersecurity

Detect review manipulation by leveraging reviewer historical stylometrics in amazon, yelp, facebook and google reviews

Nafiz Sadman, Kishor Datta Gupta, Ariful Haque, Subash Poudyal, Sajib Sen

Proceedings of the 2020 the 6th International Conference on E-Business and Applications · pp. 42–47

NLPCybersecurity

“can nlp techniques be utilized as a reliable tool for medical science?”-building a nlp framework to classify medical reports

Nafiz Sadman, Sumaiya Tasneem, Ariful Haque, Md Maminur Islam, Md Manjurul Ahsan, Kishor Datta Gupta

2020 11th IEEE Annual Information Technology, Electronics and Mobile Communication Conference (IEMCON) · pp. 0159–0166

NLPHealthcare

2019

Intrinsic evaluation of bangla word embeddings

Nafiz Sadman, Akib Sadmanee, Md Iftekhar Tanveer, Md Ashraful Amin, Amin Ahsan Ali

2019 International Conference on Bangla Speech and Language Processing (ICBSLP) · pp. 1–5

NLP