IECBES 2022 - Conference Highlights

👋 Hi there. Welcome back to my page. This week, I had an opportunity to travel to Malaysia for the first time to physically attend and present at a reputed scientific conference, IECBES 2022. That was a memorable event in my personal journey. Today, I will briefly summarize conference speeches and presentations in this highlights blog post.

1. About IECBES 2022

The IECBES stands for IEEE-EMBS Conference on Biomedical Engineering and Sciences, which is organized once every 2 years by The IEEE Engineering in Medicine and Biology Society (IEEE-EMBS) Malaysia Chapter. The 7th IECBES with the major theme of “Healthcare Personalisation: For the Future & Beyond” was held in Kuala Lumpur from December 7th to 9th December 2022.

Consistent with the theme of “Healthcare Personalisation: For the Future & Beyond”, the conference provided 6 keynote lectures and 7 invited speeches from leading academic scientists, along with 75 accepted papers, categorized into 11 tracks and one special session. Figure 1 illustrates the number of papers in each track.

Figure 1. Number of Accepted Papers in Each Track.

Insight 1: As we can observe from Figure 1, Biomedical Signal Processing is the most attractive track, followed by Internet of Things in Biomedical Engineering and Healthcare, Biomedical Imaging and Image Processing, and Biomedical Instrumentation and Devices. Might be because we overcame the pandemic, the track Pre & Post COVID-19 Pandemic Response is less attractive and only has 2 accepted papers.

Next, let’s use a word cloud visualization of the titles of all accepted papers to grasp some ideas about these works. In Figure 2, the size of each word indicates its frequency.

Figure 2. Number of Accepted Papers in Each Track.

Insight 2: From the above word cloud visualization, we can draw some comments:

• Classification, detection (actually also classification), and analysis are the most frequent tasks conducted.
• The big size of words such as “deep learning”, or “neural network” shows the broad application of AI.
• Images and signals are the most frequent data modalities used.
• The small size of the word “device” indicates that there are not many works focused on hardware.

2. Outstanding Papers

In this section, I will briefly introduce some outstanding papers that I found considerable. These papers also are in tracks I am interested in such as Biomedical Imaging and Image Processing, Biomedical Signal Processing, Cardiovascular and Respiratory Systems, Pre & Post COVID-19 Pandemic Response, and Special Session - Trends in Smarter Healthcare: AI for Images. Some of these papers have been recognized as the Top 10 best papers.

Reconstruction of Fetal Head Surface from A Few 2D Ultrasound Images Tracked in 3D Space
by Sandra Marcadent; Johann Hêches; Julien Favre; David Desseauve; Jean-Philippe Thiran
in Biomedical Imaging and Image Processing

In this pilot study, we present a novel method to reconstruct the fetal head surface, from a small set of tracked 2D ultrasound images around the transthalamic brain plane. Indeed, 3D visualization of the fetus’s prominent skull at the beginning of birth could help the obstetrician in decision-making to overcome dystocia, a delivery complication that results in labor obstruction. The use of 2D ultrasound images tracked in 3D would allow superimposing of the fetal head model to other reconstructed organs. However, fetal motion may affect the consistency of ultrasound images, in particular, if many frames are needed. Moreover, the fetal head is large at late pregnancy stages which causes occlusions in the ultrasound images. We thus propose and compare the performance of two different methods to reconstruct a fetal head surface from only 10 focused frames. Our best method achieves 1.6 mm of average reconstruction error in simulation based on an MRI dataset of 7 patients at 34-36 weeks of pregnancy.

Vector-Quantized Zero-Delay Deep Autoencoders for The Compression of Electrical Stimulation Patterns of Cochlear Implants Using STOI
by Reemt Hinrichs; Felix Ortmann; Jörn Ostermann
in Biomedical Signal Processing

Cochlear implants (CIs) are battery-powered, surgically implanted hearing aids capable of restoring a sense of hearing in people suffering from moderate to profound hearing loss. Wireless transmission of audio from or to signal processors of CIs can be used to improve speech understanding and localization of CI users. Data compression algorithms can be used to conserve battery power in this wireless transmission. However, very low latency is a strict requirement, limiting severely the available source coding algorithms. Previously, instead of coding the audio, coding the electrical stimulation patterns of CIs was proposed to optimize the trade-off between bit rate, latency, and quality. In this work, a zero-delay deep autoencoder (DAE) for the coding of the electrical stimulation patterns of CIs is proposed. Combining for the first time bayesian optimization with numerically approximated gradients of anon-differential speech intelligibility measure for CIs, the short-time intelligibility measure (STOI), an optimized DAE architecture was found and trained that achieved equal or superior speech understanding at zero delays, outperforming well-known audio codecs. The DAE achieved reference vocoder STOI scores at 13.5 kbit/s compared to 33.6 kbit/s for Opus and 24.5 kbit/s for AMR-WB.

Performance of A Wireless Electrocardiogram System Based on Wi-Fi and BLE Technology
by Nusrat Hassan Khan; S M Nafiul Hasan Joy; Fauzan Khairi Che Harun; Weng Howe Chan; Nurul Ashikin Abdul-Kadir; Moey Keith
in Biomedical Instrumentation and Devices

Wearable electrocardiogram (ECG) systems have increasingly been used in everyday life, breaking down the barriers that formerly existed only within hospitals. They allow for non-invasive continuous monitoring of a variety of heart parameters. The aim of this work is to investigate and assess the development of a user-friendly, mobile, and compact wearable ECG system for instantaneous recording. The work also presented the design of the ECG system with Autodesk EAGLE and Fusion 360 which has wireless connectivity via Bluetooth and Wi-Fi. The functionality of this ECG system is aided by the BMD101 cardio chip device, which is composed of an amplifier, filter, and 16-bit analog-to-digital converter. The results indicated a regular cardiac rhythm of 60 beats per minute (bpm), 120 bpm, and 180 bpm, respectively, along with the abnormal heart condition of ventricular tachycardia. Eventually, this study concluded with a list of key remaining obstacles as well as the potential for development in terms of result display and system software, both of which are vital for continued advancement.

Enhancing Deep Learning-based 3-lead ECG Classification with Heartbeat Counting and Demographic Data Integration
by Khiem Le; Huy Hieu Pham; Thao Nguyen; Tu Nguyen; Cuong Do; Tien Ngoc Thanh
in Cardiovascular and Respiratory Systems

Nowadays, an increasing number of people are being diagnosed with cardiovascular diseases (CVDs), the leading cause of death globally. The gold standard for identifying these heart problems is via electrocardiogram (ECG). The standard 12-lead ECG is widely used in clinical practice and the majority of current research. However, using a lower number of leads can make ECG more pervasive as it can be integrated with portable or wearable devices. This article introduces two novel techniques to improve the performance of the current deep learning system for 3-lead ECG classification, making it comparable with models that are trained using standard 12-lead ECG. Specifically, we propose a multi-task learning scheme in the form of the number of heartbeats regression and an effective mechanism to integrate patient demographic data into the system. With these two advancements, we got classification performance in terms of F1 scores of 0.9796 and 0.8140 on two large-scale ECG datasets, i.e., Chapman and CPSC-2018, respectively, which surpassed current state-of-the-art ECG classification methods, even those trained on 12-lead data.

Image-To-Graph Transformation via Superpixel Clustering to Build Nodes in Deep Learning for Graph
by Hong Seng Gan; Muhammad Hanif Ramlee; Asnida Abdul Wahab; Wan MahaniHafizah Wan Mahmud; De Rosal Ignatius Moses Setiadi
in Special Session - Trends in Smarter Healthcare: AI for Images

In recent years, convolutional neural networks (CNN) becomes the mainstream image-processing technique for numerous medical imaging tasks such as segmentation, classification, and detection. Nonetheless, CNN is limited to processing fixed-size input and demonstrates low generalizability to unseen features. Graph deep learning adopts graph concepts and properties to capture rich information from complex data structures. Graphs can effectively analyze the pairwise relationship between the target entities. Implementation of graph deep learning in medical imaging requires the conversion of grid-like image structure into a graph representation. To date, the conversion mechanism remains underexplored. In this work, image-to-graph conversion via clustering has been proposed. Locally grouped homogeneous pixels have been grouped into a superpixel, which can be identified as a node. Simple linear iterative clustering (SLIC) emerged as the suitable clustering technique to build superpixels as nodes for subsequent graph deep learning computation. The method was validated on the knee, cell, and membrane image datasets. SLIC has reported a Rand score of 0.92±0.015 and a Silhouette coefficient of 0.85±0.02 for the cell dataset, 0.62±0.02 (Rand score) and 0.61±0.07 (Silhouette coefficient) for the membrane dataset, and 0.82±0.025 (Rand score) and 0.67±0.02 (Silhouette coefficient) for knee dataset. Future works will investigate the performance of superpixel with enforcing connectivity as the prerequisite to develop graph deep learning for medical image segmentation.

3. Closing

IECBES is a rising conference and more and more attractive in the field of Biomedical Engineering and Sciences. However, the conference is still fledgling with an h-index of around 7, therefore, its accepted papers are not really high-quality. I hope the conference will continue to grow, attracting more outstanding research from all over the world in the next years.

Stay tuned for more content …