Sheraz Khan’s Post

🌟 Exciting news! Our article "A Machine-Learning-Based Motor and Cognitive Assessment Tool Using In-Game Data from the GAME2AWE Platform" has been chosen as an "Editor's Choice" in a leading informatics journal. Special thanks to my PhD supervisor, Prof. Christos Goumopoulos, for his guidance and support throughout this project. Check it out here: https://lnkd.in/dv38cfNa #MDPI #Infomatics #SeriousGames #MachineLearning #ResearchImpact #EditorsChoice

Assembling 512 EEG channels with the g.Pangolin system to get perfect spatial resolution of the visual cortex 🥰🥰. During my PhD I need 60 minutes for 64 channels. This would be 8 hours for 512 passive EEG electrodes 🥲🥲. With the g.Pangolin system we need just 30 minutes 🥰🥰🥰. 👉 This is a whole preparation video for VEP experiments that g.tec is running to show the spatial extend and boundaries of different images on the human scalp. The experiment is done with the 1024 ultra-high density g.Pangolin system in combination with g.HIamp and g.HIsys Professional coming with g.Suite 2020. 👉 g.HIamp is a 24 Bit biosginal amplifier with 256 ADCs for a perfect synchronization of all EEG channels. It can be stacked to record from 1024 channels. Each device has 4 processors included to perform a very high oversampling and ultra-steep anti-aliasing filtering to increase the signal to noise ratio. 👉 g.HIsys Professional is a MATLAB/Simulink toolbox that comes with hundreds of dedicated signal processing bocks for EEG/EP and ECoG experiments. Within a few minutes new experiments can be defined. The toolbox is extremely fast and allows to process all the 1024 channels. 👉 g.Pangolin grid electrodes have a center to center distance of only 8.6 mm and a very small diameter. This allows to record with very high spatial resolution. In contrast standard EEG electrodes have a diameter of about 8 mm and the gel under the electrodes covers a pretty large area. This limits the possible spatial resolution. Furthermore, it is not possible to bring standard electrodes close together. 👉 With this setup we can duplicate every single EEG experiment and just get better resolution. 👉Joint the BCI & Neurotechnology Spring School 2024 to see the experiment happening: https://lnkd.in/dF4pnc-u 👉 All made possible by an outstanding team: Leonhard Schreiner Matteo La Rosa Pauline Schomaker 👉 To learn how BCI technology works for neurorehabilitation: https://lnkd.in/dVDGJaRX 👉 To learn more about BCI technology in general: https://lnkd.in/ghXG2Pq6 #gtecmedical #EEG #ERP #EP

𝗣𝗿𝗼𝘃𝗶𝗱𝗶𝗻𝗴 𝗰𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝘀𝘂𝗽𝗽𝗼𝗿𝘁 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗲𝗹𝗱𝗲𝗿𝗹𝘆💡🤖 It was about seven years ago when Professor Chuan Wu from the Department of Computer Science began researching on a smart walker for the elderly, following her father’s diagnosis with Parkinson's disease. Indeed, not just her but many others understand the serious mobility issue suffered by the elderly, which undermines their quality of life. The walker developed by Professor Wu with the help of students and another professor from the Faculty of Engineering has proven to be a valuable tool. In 2023, it won a 🎖 Bronze Medal in the 48th International Exhibition of Inventions in Geneva, Switzerland, in recognition of its significant contribution in the field of mobility and elderly care. Putting her knowledge and strengths to good use Given her computer science background, Professor Wu decided to leverage her expertise and knowledge to create a device that could help the elderly move around safely and easily, both indoor and outdoor. The invention – 𝒕𝒉𝒆 𝑺𝒎𝒂𝒓𝒕 𝑹𝒐𝒃𝒐𝒕𝒊𝒄 𝑾𝒂𝒍𝒌𝒆𝒓 – combines advanced Artificial Intelligence (AI) with a home-friendly design, creating a cosy companion for seniors to stay healthy, active, and independent. Read the full article here🔗 https://lnkd.in/gEwsX_ir #elderlysupport #AIforgood #SmartRoboticWalker #HKUEngineering #ComputerEngineering

Hey #bci fellows, Please find the latest update of our #dataset Physionet EEGMMIDB in #MATLAB structure and #CSV files to leverage accessibility and exploitation https://lnkd.in/d4nf_MEw For the full dataset description and reproducibility guide, you can read https://lnkd.in/dUxnQAus Enjoy reading and exploring one of the largest #EEG datasets for #motor #imagery and motor #execution.

Machine learning with electrocardiograms: A call for guidelines and best practices for ‘stress testing’ algorithms   A potential opportunity for the ECG community is to provide guidelines to researchers to help guide ECG ML (Machine Learning) practices. For example, whilst general ML guidelines exist, there is perhaps a need to recommend approaches for ‘stress testing’ and evaluating ML algorithms for ECG analysis, e.g. testing the algorithm with noisy ECGs and ECGs acquired using common lead and electrode misplacements. This paper provides a primer on ECG ML and discusses some of the key challenges and opportunities. Bond, R., Finlay, D. D., Al-Zaiti, S. S., & Macfarlane, P. W. (2021). Machine learning with electrocardiograms: A call for guidelines and best practices for ‘stress testing’ algorithms. Journal of Electrocardiology, 69, 1–6. https://lnkd.in/dkjT5_PR. #btlindia #engineeringcareers #medicalequipment #innovation #cardiology #machinelearning #algorithms #greatplacetowork #wellhsrcertified

Atrial fibrillation (AF) is the most common cardiac arrhythmia (irregular heartbeat) affecting senior citizens, with approximately 8% of people over the age of 65 experiencing some degree of AF. The relative frequency of these episodes is a reliable predictor for future hospitalizations due to heart failure and stroke, so early detection of such arrhythmias is essential to improved health outcomes. Senior citizens from underserved and underprivileged populations tend to have limited access to health care facilities and transportation, so remote monitoring of these patients using wearable devices (e.g. smartwatches) could drastically improve their quality of life. This project will focus on designing an automated classifier capable of distinguishing subjects with normal heartbeats from those experiencing AF. Possible techniques include Shannon entropy, k-nearest neighbors, and support vector machines, along with various signal processing methods. Interested in working on an AI Project on ‘Detecting Cardiac Arrhythmias’? Thinking of pursuing Data Science, Engineering or Pre-Med Tracks in College? STEAM in AIᵀᴹ CAN BE YOUR DIFFERENTIATING FACTOR!! Winter 2023 Session starts December 20th. Register Today! #STEAMinAI #Collegeprep #AIskills #BreakingintoAI #datascience #engineering #premed #collegeapplications #collegeadmissions #highschoolstudents

🎉 This week marked the end of my first prize competition on Kaggle, where our team got a silver medal (#53 out of ~2700 teams) 😋. The goal of this competition was to model the ability of experts to classify 1 of 6 harmful brain states (seizure subtypes) from EEG data provided by Harvard Medical School. We tried different neuroscience-inspired approaches (i.e. cleaning signal, channel rejection+topographic interpolation, ERP-like data extraction etc.), but in the end, our top-performing solution was to feed stacked time-frequency spectrogramms into a set of large pre-trained models. Now I know why people in the ML/DL community recommend beginners to start Kaggle to gain experience :) I learned so many practical tricks from public kernels and from my teammate Sergey Sakharovskiy, which they don't teach even in the top ML/DL courses and/or books. I hope that the winning solutions could help to improve the diagnosis of seizure subtypes and thus improve treatment. Thanks, Kaggle, HMS, and the team for this opportunity! Link to competition: https://lnkd.in/eHhWfvY7

Our paper, titled "Self-assessment of Affect-Related Events for Physiological Data Collection in the Wild Based on Appraisal Theories"  has just been published. In this study, we introduce a novel technique for creating datasets for affective computing which is based on appraisal theories You can find more information here: https://lnkd.in/dgdYUANs #DIBRIS #affectivecomputing

Excited to present our work at NeurIPS next week. Looking forward to connecting with fellow attendees, and very open to discussing post-PhD opportunities. 🔎 Self-Supervised Learning (SSL) is all about learning how data varies when you don't have labels. Our focus is learning from multiple perspectives or 'views' of the same object or scene. For instance, consider two different photographs 📸 of the same object from different angles 🐕/ 🐶 , or a video and corresponding audio clip 🎞/🎶 . The goal is to learn embeddings of these different views in such a way that: 1) The embeddings of different views of the same object are similar to each other to capture the essence of the object, regardless of the angle or modality (like visual or auditory). 🐶✅🐕 2) These embeddings are distinct from those of different objects, which is crucial for the model to understand that different objects or scenes should have different representations. 🐶❌😺 To solve the problem, we revisit the roots of classical Canonical Correlation Analysis (CCA) in a stochastic setting to achieve orthogonal embeddings that highlight diverse aspects of variance in objects. The result? A 3️⃣ Line loss function which plugs into your favourite self-supervised pipeline (Lightly, Solo-Learn), and matches the state-of-the-art without any tuning 🏆 . For more info, see: 👩💻 Our code: https://lnkd.in/eijr_wEH 📜 Our paper: https://lnkd.in/eeTetGk3 Thank you to my co-author Lennie Wells for turning my engineering tricks into rigorous maths! Thank you also to G-Research for their research grant which has made my travel possible, and to the EPSRC-funded UCL Centre for Doctoral Training in Intelligent, Integrated Imaging in Healthcare (i4health). #neurips2023 #SelfSupervisedLearning

🚀 COME GET YOUR GUARDIAN 🚀 We are thrilled to announce the launch of our automated checkout page where you can now purchase the IDUN Guardian directly from our site. Visit us at https://lnkd.in/ekiPfWd3 🧠 Why choose the IDUN Guardian? ✅ Raw Data Access: Gain insights with immediate access to raw EEG data. ✅ Python SDK: Real-time data access and the ability to develop personalized interventions right at your fingertips. ✅ Console Webapp: An intuitive platform for recording, accessing, and managing your EEG data efficiently. ✅ Powerful API: Seamlessly integrate with your existing systems to enhance functionality and data flow. ✅ Lab Streaming Layer (LSL): Connect and synchronize multiple devices for complex experimental setups. ✅ Advanced Classifiers: Utilize real-time prediction capabilities for jaw clenching, frequency bands, and eye movement. 👩💻👨💻 For developers, researchers, and innovators looking to harness the power of neurotechnology—IDUN Guardian is your perfect partner. Transform the way you interact with technology. Purchase your IDUN Guardian today and start exploring the limitless possibilities of EEG! #IDUNTechnologies #Neurotechnology #EEG #BrainComputerInterface #Innovation #BCI #NeuroTechX