Shovito Barua Soumma

Invited talk: Wearable AI Systems for Health Promotion and Disease Management

Thu, 30 Apr 2026 15:00:00 -0700

Hybrid Attention Model Using Feature Decomposition and Knowledge Distillation for Blood Glucose Forecasting

Fri, 10 Apr 2026 08:01:35 -0700

Self-Supervised Learning and Opportunistic Inference for Continuous Monitoring of Freezing of Gait in Parkinson’s Disease

Tue, 24 Mar 2026 08:01:35 -0700

Stress Mindset Matters: Rethinking Mental Stress Detection with Multimodal Wearable Sensors

Wed, 18 Mar 2026 12:00:00 -0700

Trustworthy AI in Digital Health: A Comprehensive Review of Robustness and Explainability

Fri, 06 Mar 2026 06:59:00 -0700

Large language model agents for biological intelligence across genomics, proteomics, spatial biology, and biomedicine

Tue, 10 Feb 2026 08:01:35 -0700

From Indicators to Insights: Diversity-Optimized for Medical Series-Text Decoding via LLMs

Wed, 21 Jan 2026 12:00:00 -0700

A large sensor foundation model pretrained on continuous glucose monitor data for diabetes management

Wed, 22 Oct 2025 12:00:00 -0700

AI-Powered Wearable Sensors for Health Monitoring and Clinical Decision Making

Wed, 08 Oct 2025 16:30:00 -0700

SenseCF: LLM-Prompted Counterfactuals for Intervention and Sensor Data Augmentation

Mon, 08 Sep 2025 16:30:00 -0700

Digital-twin based optimization of bolus insulin dosing in pediatric type 1 diabetes: an in silico feasibility study

Wed, 23 Jul 2025 12:00:00 -0700

Detection and Severity Assessment of Parkinson’s Disease by Analysis of Wearable Sensors Data Using Gramian Angular Fields and Deep Convolutional Neural Networks

Mon, 26 May 2025 01:01:35 -0700

Freezing of Gait Detection Using Gramian Angular Fields and Federated Learning from Wearable Sensors

Wed, 09 Apr 2025 16:30:00 -0700

Design and Implementation of a Scalable Clinical Data Warehouse for Resource-Constrained Healthcare Systems

Wed, 09 Apr 2025 16:29:00 -0700

Enhancing Metabolic Syndrome Prediction with Hybrid Data Balancing and Counterfactuals

Wed, 09 Apr 2025 16:20:00 -0700

LLM-CGM: A Benchmark for Large Language Model-Enabled Querying of Continuous Glucose Monitoring Data for Conversational Diabetes Management

Wed, 12 Feb 2025 12:00:00 -0700

ConvBoost: Boosting ConvNets for Sensor-based Activity Recognition

Tue, 03 Dec 2024 12:00:00 -0700

Domain-Informed Label Fusion Surpasses LLMs in Free-Living Activity Classification

Wed, 30 Oct 2024 08:01:35 -0700

Time-LLM: Time Series Forecasting by Reprogramming Large Language Models

Wed, 16 Oct 2024 12:00:00 -0700

Wearable-Based Real-time Freezing of Gait Detection in Parkinson's Disease Using Self-Supervised Learning [Abstract]

Mon, 07 Oct 2024 08:01:35 -0700

LIMU-BERT: Unleashing the Potential of Unlabeled Data for IMU Sensing Applications

Wed, 11 Sep 2024 12:00:00 -0700

Health-LLM: Large Language Models for Health Prediction via Wearable Sensor Data

Wed, 17 Jul 2024 12:00:00 -0700

Teaching

Wed, 19 Jun 2024 23:58:23 -0800

BMI 504/404 Intro to Clinical Environments (Spring 26)

Instructor: Dr. Anita Murcko

BMI 502 Foundations of Biomedical Informatics Methods I (Fall 24, 25)

Instructor: Dr. Hassan Ghasemzadeh

BMI 570 Symposium (Fall'25)

Instructor: Dr. Anita Murcko

BMI 310: App Development for Clinical and Population Health (Spring 2025)

Instructor: Shovito B Soumma
Syllabus: click here

BMI 201 Introduction to Clinical Informatics (Spring 2024)

Instructor: Dr. Marcela Aliste

AI-Powered Detection of Freezing of Gait Using Wearable Sensor Data in Patients with Parkinson’s Disease [Abstract]

Fri, 31 May 2024 00:00:00 +0000

Self-Supervised Learning of Pretext-Invariant Representations

Tue, 16 Apr 2024 15:00:00 -0700

Invited talk: Mobile App & Wearable Based in Home Parkinson Disease Management

Thu, 11 Apr 2024 12:00:00 -0700

Real-Time Patient Adaptivity for Freezing of Gait Classification Through Semi-Supervised Neural Networks

Tue, 20 Feb 2024 15:00:00 -0700

Data Augmentation of Wearable Sensor Data for Parkinson’s Disease Monitoring using Convolutional Neural Networks

Wed, 20 Dec 2023 13:30:00 -0700

SimCLR: A Simple Framework for Contrastive Learning of Visual Representations

Wed, 01 Nov 2023 13:30:00 -0700

Metabolic Health

Mon, 01 May 2023 23:58:08 -0800

Metabolic health refers to the balance and optimal functioning of the body’s metabolic processes, including glucose regulation, lipid metabolism, and energy utilization. With the rise of chronic diseases like diabetes, AI-driven technologies have become crucial tools for diabetes management. These technologies leverage data science and artificial intelligence to analyze large volumes of patient data, such as continuous glucose monitoring (CGM) data, to provide personalized insights and interventions.

CGM data, collected through wearable devices (like Dexcom, Abbott’s FreeStyle Libre, Medtronic etc.), offers real-time monitoring of a patient’s blood glucose levels. AI algorithms process this data to generate actionable insights, helping individuals make informed decisions about their diet, medication, and physical activity. By analyzing patterns and trends in CGM data, AI-driven systems can provide predictive alerts for potential hyperglycemic or hypoglycemic events, enabling users to take proactive measures.

The integration of AI into diabetes management not only enhances individual control over glucose levels but also facilitates healthcare professionals' ability to provide targeted interventions. As AI technologies continue to evolve, they hold the promise of revolutionizing the way we approach metabolic health, leading to more effective and personalized strategies for diabetes management.

Machine Learning Approaches to Metastasis Bladder and Secondary Pulmonary Cancer Classification Using Gene Expression Data

Sat, 17 Dec 2022 08:01:35 -0700

UG Research

Wed, 13 Apr 2022 23:58:08 -0800

HCI, Sociology, Human Psychology
PI: Prof. Dr. A B M Alim Al Islam, CSE, BUET We are examining if the use of social media in Bangladesh contributes to a breakdown in communication between family members. Data has been collected using semi-structured face-to-face interviews. The Thematic technique will be used to collect and evaluate data from the recorded audio scripts of our interviewees in order to identify a pattern.

South Asian Public Digital Service Centers and the Risk to User Privacy (2019)

HCI, Privacy This study looked at 19 digital service centers in Bangladesh. The findings showed that customers of these centers were vulnerable to privacy breaches due to a lack of infrastructure, local power politics, a lack of knowledge, and inadequate protection mechanisms.

Android Malware Detection Based on System Calls Using NLP and Machine Learning Algorithms (2022)

ML, Security and Privacy
PI: Prof. Dr. Md Shohrab Hossain, CSE, BUET
Collaborator: A S M Ahsan-Ul Haque, SDE, Amazon.com, M.Sc, University of Virginia [2022], B.Sc in CSE, BUET [2017] I wanted to extend one of Dr. Hossain’s previous research where he used 1-step transition probability between the system calls to detect malware apk. As the prior model cannot capture the order or structure of system calls, it lacks semantic information. So after conducting a literature review to identify the limitations of the existing method for evading System Call-based Intrusion Detection System (IDS), I proposed a Machine Learning based robust dynamic method to detect malware apks, that can automatically execute the code routines as well as generate the user behavior of the android app.
Objectives & Findings:
- Generate user behavior during the system call retrieval from a virtual android device
- Use a universal sentence encoder to represent each system call with an equivalent vector of 512 dimensions.
- Finally, a random forest classifier with 100 estimators is used to ‘accurately’ classify our data. Along with the Random Forest (RF) model, we ran experiments with other models, such as Logistic Regression, Multilayer Perceptron (MLP) and XGBoost to compare our results.

Data Warehouse Design for Health Sectors and Outbreak Prediction (2021-22)

Deep Learning, Data mining and Information Systems
PI: Prof. Dr. Abu Syed Md. Latiful Haque, CSE, BUET

I have developed a clinical big data platform prototype-NCDW, integrating ambient data from 34 weather stations of Bangladesh Meteorological Department (BMD) as a proof of concept and solved the fundamental obstacle for data-driven communicable and non-communicable disease research, including record-linkage, privacy, and security, standardization, and interoperability. I submitted the conceptual design of my proposed system to Bangladesh Space Research and Remote Sensing Organization (SPARRSO) and secured their “Research Fellowship”. This platform enhance descriptive, diagnostic,predictive, and prescriptive analysis and research for a wide variety of diseases.

Main Objectives:

Estimate the size of the NCDW and facilitates regional and national decision support, intelligent disease analysis, knowledge discovery, and data-driven research
Develop a model to predict the number of cases of a given month of a given district. Forecast a disease outbreak

Published in 47th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), July 14–17, 2025, Copenhagen, Denmark [Preprint !!!]

Course Projects

Thu, 13 Jan 2022 23:58:23 -0800

Tuberculosis Detection using Chest X-ray with Deep Learning (2022)

Presentation

In this work, I have detected TB reliably from the chest X-ray images using image pre-processing, data augmentation and deep-learning classification techniques. Four different deep CNNs (ResNet18, ChexNet, InceptionV3, and MobileNet) were used for transfer learning from their pre-trained initial weights and were trained,validated and tested for classifying TB and non-TB normal cases.

Portable Heart Rate Monitoring System (2020)

L3 T1 (UG 5th Semester), GitHub
Technologies: AtMega32 micro controller, HC-05 Bluetooth Module, 16x2 LCD Display, Android Device

Portable heart rate monitor is a personal monitoring device that measures the heart rate using optical sensors in real time and send the measured the data to directly to an android device. We are developing prototype of this application using the continuous monitoring of parameters to detect and predict the heart attack and generate an alarm.

Spondon - A Medical App (2020)

GitHub, Technologies: NodeJs, MongoDB, Android

User can search a blood donor; Search parameter: name, blood group, division, district, availability User can search an ambulance or oxygen cylinder; Search parameter: name division, district User can show the live bed status of all the hospitals of Bangladesh and filter the hospitals according to hospital name, division, district, availability, last update We have an admin side built in React from where admin can perform CRUD operation for ambulance and oxygen cylinder.

Algorithm Simulator (2018)

L1 T2 (2nd semester) GitHub
Technology: JavaFX, Java, InelliJ

Aim of this project was to get familiar with OOP language, GUI designing, and simulate different algorithms visually by taking input from users. This is a java project that shows simulation of 3 sorting algorithms and BFS. JavaFx is used for the UI part. In our simulation part of sorting algorithm, we resize the bar of input given by the user so that it can fit in window.