Ending the Semester on a High Note: A Successful SLDR for BeatNet

This week marked an exciting milestone for Team BeatNet — the completion of our System-Level Design Review (SLDR) and the wrap-up of the Fall semester. Our presentation showcased how the project has evolved far beyond a standard deep learning classification task. We’ve successfully transitioned into building a Self-Supervised Foundation Model for ECG learning — positioning BeatNet at the forefront of intelligent mobile cardiology solutions.

A Big Thank You

We want to express our sincere gratitude to:

• Aventusoft, our project sponsor, for continuously challenging us with ambitious goals — especially in adapting advanced AI techniques to single-lead wearable ECG.
• Dr. Kejun Huang, our team coach, for expert guidance, thoughtful feedback, and helping us refine both our engineering approach and long-term system vision.

Your support and encouragement helped drive every breakthrough we achieved this semester.

Looking Ahead: Ready for the Spring Sprint

What’s next for BeatNet?

With the holidays around the corner, our immediate plan is simple: rest, recharge, and reconnect with family and friends. Once the Spring semester kicks off, we’ll resume at full speed — laser-focused on:

• Validating segmentation and classification performance
• Advancing explainability for clinical trust
• Model compression and deployment to mobile

The momentum is strong, and we’re excited to accelerate our journey toward a real-world, deployable system.

From Pretraining to SLDR : A New Milestone for Team BeatNet

This week, We made significant progress by moving from model development to system-level validation in preparation for our upcoming System-Level Design Review (SLDR). After weeks of developing and fine-tuning the key elements of our ECG analysis pipeline, the team achieved important technical and documentation milestones that advance us toward a deployable and clinically valuable system.

Key Accomplishments This Week

SLDR Preparation and Architecture Finalization

A significant portion of our efforts went toward preparing the SLDR report and presentation, which required consolidating architecture decisions, documenting interfaces, validating functional requirements, and aligning our work with Aventusoft’s deployment expectations. This process allowed us to refine our system blueprint and ensure each subsystem—preprocessing, fiducial detection, classification, and deployment—fits into a cohesive and verifiable design.

Progress on the Foundation Model Pipeline

We continued advancing the foundation model workflow by evaluating lightweight architectures and early-stage distillation strategies to support mobile inference. This included comparing smaller CNNs and Transformer variants, analyzing their compute profiles, and assessing whether they meet BeatNet’s <30-second runtime requirement for mobile deployment.

Fiducial Detection & Classification Integration

Work also progressed on the integration between the fiducial regression module and the arrhythmia classifier. We refined the windowing logic, updated segmentation strategies based on coach feedback, and prepared the models for the upcoming validation phase, which will begin after SLDR.

Explainability as a New Priority

Following sponsor discussions, explainability has become a major focus for the next sprint. Physicians must understand why the model makes its decisions. In line with that requirement, the team has begun planning explainability modules such as:

• Attention heatmaps for rhythm classification
• P/Q/R/S/T waveform overlays for delineation transparency
• Interpretability dashboards for both cloud and mobile inference

These additions will support clinical trust and align the system with regulatory expectations for traceability and interpretability.

Documentation & Technical Review

We completed substantial sections of the SLDR document—including system architecture, requirements traceability, software interfaces, and deployment considerations. This helped us validate the technical cohesion of the project and prepare for sponsor review.

Next Steps: Validation, Explainability, and Deployment Readiness

With SLDR approaching, our focus for next week will shift toward running validation experiments for both the segmentation and classification modules, implementing the first generation of explainability tools, and continuing our research on model compression, quantization, and distillation to support mobile deployment. At the same time, we will be finalizing the SLDR presentation materials to clearly communicate our path toward cloud and mobile integration. As we enter a phase where accuracy, interpretability, and efficiency converge, the SLDR checkpoint will serve as an important opportunity to demonstrate our progress and gather targeted feedback from our sponsors, guiding us into the next stage of system development.

See you next week!

From Pretraining to Explainability : A New Focus for BeatNet

All of our efforts this week were geared toward one major goal: validating our new foundation-model This was a landmark week for Team BeatNet! After spending the last few weeks building our core pipeline, we successfully completed our first full pretraining run of the foundation model. We presented these exciting initial results—which already show the model is learning the deep structure of ECGs—to our sponsors, Dr. Kejun Huang and Dr. Keider Hoyos.

This successful review meeting not only validated our technical approach but also gave us a critical new focus for the next sprint: clinical interpretability.

Key Accomplishments This Week

First Pretraining Round Complete We successfully completed the first round of pretraining using our contrastive learning approach. By teaching the model to recognize ECG segments from the same patient as “positive pairs,” we are forcing it to learn the fundamental, patient-invariant features of a heartbeat. Early results are promising, showing “emerging latent structure” in the embeddings, which validates our entire model setup.

Strategic Focus on Clinical Interpretability During our technical review, Dr. Hoyos emphasized that for this tool to be trusted by physicians, it cannot be a “black box.” It’s not enough for the AI to be accurate; doctors must understand why it’s making a specific diagnosis. Following this guidance, the sponsors approved our plan to integrate explainability modules (like attention heatmaps) in the next sprint.

Full Pipeline Integration and Fine-Tuning Initiated With the foundation model prototype finalized (combining CNN encoders with a Transformer decoder), we have begun attacking the downstream tasks. We have already started the first fine-tuning experiments for arrhythmia classification and have implemented the pipeline for fiducial landmark (P/Q/R/S/T) regression.

Deployment & Optimization Research In parallel, we have been optimizing the model and preparing for deployment. We worked on benchmarking different window lengths (5s vs. 10s) to find the sweet spot for context learning. We also continued research on model distillation and lightweight architectures, assessing how we can shrink this powerful model to run efficiently on BeatNet’s embedded system.

Next Steps: Validation, Explainability, and Deployment

With the pretrained model in hand, next week is all about validation and implementation. We will begin cross-validation testing, using the model’s embeddings for both arrhythmia classification and fiducial detection.

A major priority will be implementing the new explainability modules—generating waveform attention maps and fiducial localization overlays. We will also evaluate model compression and quantization strategies for mobile deployment. Finally, all of this will be packaged into our updated slides for the upcoming System-Level Design Review (SLDR), where we will highlight our model’s explainability and a clear path to deployment.

See you next week!

From Public Validation to Pipeline Implementation

All of our efforts this week were geared toward one major goal: validating our new foundation-model strategy with the wider UF community and beginning the core implementation of our self-supervised pipeline. After a successful presentation at the Prototype Inspection Day (PID), Team BeatNet is now fully focused on building the components necessary to pre-train our model on large-scale datasets.

This week was about turning plans into concrete action—coding the contrastive learning framework, benchmarking encoder backbones, and preparing scripts for downstream fine-tuning.

Key Accomplishments This Week

Foundation-Model Strategy Defined:Successful Prototype Inspection Day (PID)
We presented our prototype progress and foundation-model strategy to UF faculty and alumni. Judges praised the team’s clear communication and cohesive technical direction. Dr. Chenhao Wang highlighted our clarity, Dr. Tingsao Xiao called our foundation-model approach “intuitive and reasonable,” and Dr. Catia Silva encouraged us to begin embedding implementation as soon as possible.

Contrastive Learning Pipeline Initiated
We began implementing the SimCLR-style contrastive learning pipeline, developing an augmentation module that generates positive pairs through noise, jitter, and scaling—an essential step toward robust patient-invariant embeddings.

Downstream Pipeline Development
In parallel, we started building the downstream fine-tuning workflow to adapt pretrained embeddings for arrhythmia classification and later for fiducial landmark regression.

Encoder Benchmarking
We benchmarked multiple 1D encoder backbones—including ResNet1D and EfficientNet1D—to identify the optimal architecture for feature extraction efficiency and downstream transferability.

Improved Communication Flow
Following judge feedback, the team is also refining presentation visuals, ensuring balanced speaking roles, and better illustrating the link between preprocessing, embedding, and disease classification.

Next Steps: Pre-training and Visualization

Next week, we will execute the first pre-training runs on the PTB-XL dataset, finalizing and debugging the contrastive learning script. We will develop a t-SNE visualization notebook to evaluate whether embeddings effectively cluster normal and arrhythmic segments, and improve visualization of time-positional encoding within our Transformer prototype.

We will also begin preparing documentation for the upcoming System Level Design Review (SLDR) and conduct internal comparisons between baseline and foundation-model performance.

See you next week!

Advancing the Foundation Model

This week was a significant milestone for us as we moved from architectural design to actively prototyping our ECG foundation model. Following last week’s design discussions, we concentrated on developing and testing the initial version of our self-supervised pretraining pipeline, turning the idea of contrastive learning from theory into practice.

Our collective goal was to transform unlabeled ECG data into structured, patient-invariant representations that can serve as the backbone for future landmark detection and disease classification models. The week was defined by rigorous experimentation, cross-validation, and early visualization of emerging cardiac signal embeddings.

Key Accomplishments This Week

Foundation-Model Strategy Defined

We conducted a detailed technical meeting with Dr. Kejun Huang and Dr. Keider Hoyos and finalized the move toward a foundation model built on large-scale ECG corpora such as PTB-XL and MIMIC-III/IV. The model will leverage contrastive pre-training, treating 5-second and 10-second ECG windows from the same patient as positive pairs, while windows from different patients act as negative pairs. This approach allows the model to learn invariant, patient-specific embeddings.

Explored Alternate Self-Supervised Approaches

We evaluated potential pre-training routes including masked autoencoder (MAE) learning and distance-map regression, which predicts landmark-wise distance functions rather than explicit waveform peaks.

Semester 2 Modeling Pillars Finalized

Our Semester 2 (Spring 2026) deliverables are now centered around three primary components:

Landmark Detection Model: multi-output regression to identify P/Q/R/S/T time-stamps.

Arrhythmia & Disease Classification: fine-tuning foundation embeddings for AFib, PVC, LBBB/RBBB, and conduction block detection.

Model Distillation for Mobile Deployment: quantization and pruning to adapt the foundation model for Aventusoft’s single-lead (500 Hz) device.

Single-Lead Adaptation Framework

Discussions also clarified strategies for retraining 12-lead models to function effectively with single-lead input, aligning with Aventusoft’s mobile ECG system requirements.

Next Steps: Prototype Implementation

In the coming week, our focus will shift toward implementing the prototype foundation-model pipeline. We will begin with contrastive pre-training using PTB-XL segments to establish the base embedding space and develop an augmentation module capable of simulating inter-patient variability through signal inversion, noise injection, and time-warping. Next, we will visualize these learned embeddings using t-SNE to validate whether the model can effectively cluster normal and arrhythmic patterns. Simultaneously, we plan to benchmark different encoder backbones, such as ResNet and EfficientNet-1D, to identify the most efficient architecture for downstream fine-tuning. Finally, our team will document the complete workflow from pre-training to distillation for inclusion in the upcoming System Level Design Review (SLDR) and coordinate with Aventusoft regarding access to internal anonymized ECG data and available GPU resources.

This week was about transforming technical insight into implementation readiness—laying the foundation for self-supervised ECG intelligence that bridges research innovation with real-world device deployment.

See you next week!

Laying the Groundwork for the Foundation Model

All of our efforts this week are geared toward one major goal: advancing our project into the foundation-model implementation phase following our successful concept validation. After a productive technical meeting with Dr. Kejun Huang and Dr. Keider Hoyos, Team BeatNet solidified our strategy to move beyond basic CNN/UNet baselines and begin building a self-supervised ECG foundation model using large-scale datasets such as PTB-XL and MIMIC-III/IV.

This week was about turning plans into concrete action—defining how contrastive pre-training, masked autoencoder learning, and distance-map regression will shape the next generation of ECG analytics for Aventusoft’s BeatNet device.

Key Accomplishments This Week

Foundation-Model Strategy Defined:
We finalized the move toward a foundation model built on large-scale ECG corpora such as PTB-XL and MIMIC-III/IV. The model will leverage contrastive pre-training, treating 5-second and 10-second ECG windows from the same patient as positive pairs, while windows from different patients act as negative pairs. This approach allows the model to learn invariant, patient-specific embeddings robust to signal variations.

Explored Alternate Self-Supervised Approaches:
We evaluated potential pre-training routes including masked autoencoder (MAE) learning and distance-map regression, which predicts landmark-wise distance functions rather than explicit waveform peaks—broadening the range of modeling techniques under consideration.

Semester 2 Modeling Pillars Finalized:
Our Semester 2 deliverables are now centered around three primary components:

1. Landmark Detection Model – multi-output regression to identify P/Q/R/S/T time-stamps.
2. Arrhythmia & Disease Classification – fine-tuning foundation embeddings for AFib, PVC, LBBB/RBBB, and conduction block detection.
3. Model Distillation for Mobile Deployment – quantization and pruning to adapt the foundation model for Aventusoft’s BeatNet single-lead (3-electrode, 500 Hz) device.

Single-Lead Adaptation Framework:
Discussions also clarified strategies for retraining 12-lead models to function effectively with single-lead input, aligning with Aventusoft’s mobile ECG system requirements.

Next Steps: Prototype Implementation

In the coming week, our focus will shift toward implementing the prototype foundation-model pipeline. We will begin with contrastive pre-training using PTB-XL segments to establish the base embedding space and develop an augmentation module capable of simulating inter-patient variability through signal inversion, noise injection, and time-warping. Next, we will visualize these learned embeddings using t-SNE to validate whether the model can effectively cluster normal and arrhythmic patterns. Simultaneously, we plan to benchmark different encoder backbones, such as ResNet and EfficientNet-1D, to identify the most efficient architecture for downstream fine-tuning. Finally, our team will document the complete workflow from pre-training to distillation for inclusion in the upcoming System Level Design Review (SLDR) and coordinate with Aventusoft regarding access to internal anonymized ECG data and available GPU resources.

This week was about transforming technical insight into implementation readiness—laying the foundation for self-supervised ECG intelligence that bridges research innovation with real-world device deployment.

See you next week!

A Successful PDR : Moving from Plan to Prototype

This week was a major milestone for our team as we successfully presented our Preliminary Design Review (PDR) to our sponsor, Aventusoft. The presentation was the culmination of weeks of intensive data exploration and planning, and the positive feedback we received has given us a strong mandate to move forward into the implementation phase.A Successful PDR: Moving from Plan to Prototype

Key Accomplishments This Week

Our focus this week was squarely on delivering a professional and technically sound PDR. The main highlights include:

• Successful PDR Presentation: We delivered our formal PDR presentation to our sponsor, liaisons, and faculty coach. The session was a success, and we received valuable feedback confirming that our technical plan, architecture, and project roadmap are aligned with sponsor expectations.

• Finalized Technical Strategy: The PDR process helped us solidify our technical strategy for the coming weeks. We confirmed our plan to begin with a baseline CNN model, tackle class imbalance with a weighted loss function, and benchmark performance using metrics like AUROC and F1-Score.

• Incorporating Sponsor Feedback: We received constructive feedback during the PDR, particularly regarding the need for a robust preprocessing pipeline and a clear plan for model explainability. This guidance has been integrated into our immediate action plan.

Next Steps: Building the Foundation

With a successful PDR now behind us, our efforts this week are geared towards turning our plan into a functional prototype. Our immediate tasks are to:

1.Finalize all PDR documentation based on the feedback and secure sponsor sign-off.

2.Begin the hands-on implementation of our data preprocessing pipeline, including the filtering, normalization, and segmentation modules.

3.Start the development of our baseline model and establish the training loop on the public datasets.

This week was about validating our strategy and earning the confidence of our sponsor. Now, the real engineering work begins, and we are excited to start building the core components of our ECG analysis system.

See you next week!

Getting up for the Sponsor Review: Insights from Data and Peer Feedback

This week was a critical period of refinement and preparation for our team. We focused on two parallel efforts: conducting our first deep dive into the public ECG datasets and, just as importantly, processing the valuable feedback from our in-class PDR peer review session. This combination of hands-on data exploration and constructive criticism has significantly sharpened our focus as we prepare for our formal presentation to our sponsor, Aventusoft.

Key Accomplishments This Week

Our progress this week was driven by both technical analysis and strategic refinement. The main highlights include:

Coordinated Exploratory Data Analysis (EDA): Our team completed a coordinated first-pass analysis of five key public datasets: PTB-XL, MIT-BIH, the CPSC 2020 Challenge dataset, LUDB, and the Chapman-Shaoxing dataset. This effort allowed us to visualize the signals, confirm technical specifications, and identify available labels for our target conditions.

Uncovering a Universal Challenge: Class Imbalance: A crucial finding from our EDA is the presence of severe class imbalance across all datasets. This insight is vital, as it confirms that addressing this imbalance through techniques like class weighting or resampling will be a core part of our modeling strategy.

Incorporating Peer Review Feedback: After presenting our draft PDR in a peer review session this week, we have been actively incorporating the constructive feedback received. Our action plan focuses on improving our presentation pacing and refining the clarity of our technical explanations to ensure our message is clear and impactful.

Streamlining Collaboration: To enhance our workflow, we have successfully established a team GitHub repository. This provides a central platform for code sharing and version control as we move into the implementation phase.

Next Steps: The Formal Sponsor PDR

All of our efforts this week are geared towards one major goal: the formal Preliminary Design Review (PDR) with our sponsor, Aventusoft, scheduled for this coming Tuesday, October 14th. Our immediate tasks are to apply the peer feedback to finalize both the presentation slides and the formal PDR report. Immediately following a successful PDR, we will begin the hands-on implementation of our data preprocessing pipeline and our baseline CNN model.

This week was about turning plans into action and feedback into improvement. The combination of insights from real-world data and our peers has left us better prepared and more confident for our upcoming sponsor review. We look forward to presenting our finalized plan and initial findings.

See you next week!

From Strategy to Practice: Key Technical Guidance from This Week’s Meeting

This week, our team held a productive meeting with our sponsor liaison engineers, which not only reinforced our project objectives but also offered essential technical advice on particular techniques for data processing and model development.

Fine-Tuning Our Technical Strategy & New Resources

We reaffirmed our two core missions: Fiducial Landmark Detection and Arrhythmia Classification. More importantly, we received several key implementation details during the meeting:

Access to Internal Data: An exciting update is that Aventusoft will provide us with an internal dataset of “normal” ECGs from company employees. This will allow us to get familiar with their device’s signal characteristics early on, preparing us for future model transfer.

Fiducial Landmark Detection as a Regression Problem : Dr. Hoyos clarified that the task of detecting PQRST points should fundamentally be treated as a regression problem, with the goal of predicting the precise sample location of the waveform peaks, rather than as a classification problem. Aventusoft’s standard approach is to first detect the R-peak, use it to center and extract a single heartbeat, and then train a multi-output regressor to predict the relative locations of the other points.

Strategy for Using Multi-Lead Data: For small datasets like LUDB (with only 200 recordings), we should not use all 12 leads to augment the data. Because the morphology of some leads (e.g., an inverted R-peak) differs too much from our target device’s signal, it could confuse the model. The guidance is to select the 3-4 leads that are most morphologically similar to Lead I and Lead II.

The Necessity of Data Augmentation: For small datasets, data augmentation is crucial for success. We were encouraged to research methods like adding noise or applying time-stretching/shrinking to the signal, while ensuring that the annotation positions are transformed accordingly.

Next Steps: A Data Deep Dive with New Guidance

With this clearer technical focus, our data exploration next week will be more precise. As each team member examines their assigned dataset, besides the usual exploratory data analysis tasks, we will concentrate on identifying shared disease labels across the various datasets, such as LBBB, to assess the possibility of merging these databases to create a larger training set in the future. Simultaneously, we will emphasize conducting a comprehensive literature review before starting model development from scratch, looking for and referencing existing studies that have already worked with these datasets so we can leverage established methods. When analyzing 12-lead ECG data, we will limit our focus to Lead I, Lead II, or other leads with similar morphology, to better replicate the real-world conditions in which our final product—a wearable device—will function.

This week’s meeting was a key step in moving us from high-level goals to concrete implementation details. We are excited to move forward with this valuable guidance and begin building the foundational components of our analysis pipeline.

See you next week!

From Strategy to Signals: Kicking Off ECG Analysis

This week, our team, BEATNET, made significant progress in defining the technical roadmap for our project. A productive meeting with our liaisons at Aventusoft provided crucial clarity and set a clear direction for the weeks ahead.

Defining Our Mission: Project Goals and Data Strategy

The primary outcome of our meeting was the confirmation of our two main project goals: arrhythmia classification (specifically targeting conditions like AFib, Flutter, and PVCs) and the detection of ECG landmarks (fiducial points). We learned that while Aventusoft has implemented Q and R point detection, the P, S, and T points remain open tasks for us to tackle.

Since most of Aventusoft’s data is currently unlabeled, we will begin by using well-known public datasets for our initial model development, including PTB-XL and MIT-BIH. This approach will allow us to build and validate our models before applying them to Aventusoft’s data in later stages.

Architecting Our Approach: Preprocessing and Deep Learning

A key technical decision from our meeting was to focus on a deep learning approach where the raw ECG signal is fed directly into the neural network. The network itself will act as a feature extractor, which avoids the need for manual feature engineering and allows the model to learn the most predictive patterns from the data.

To prepare the data for our models, we received clear guidance on the preprocessing pipeline. The core steps will include:

• Applying a Butterworth bandpass filter to clean the signal
• Resampling all data to a standard 500 Hz frequency
• Segmenting the recordings into 5 or 10-second windows for analysis
• Applying z-score normalization to standardize the signal amplitude

Our dataset exploration revealed that PTB-XL contains 21,837 clinical 12-lead ECGs from 18,885 patients with 10-second recordings and comprehensive multi-label annotations across 71 diagnostic classes. Meanwhile, MIT-BIH provides longer recordings but focuses primarily on arrhythmia detection with beat-level annotations. This diversity will strengthen our model’s generalization capabilities.

Next Steps: Diving into the Data

With a clear plan in place, our immediate focus shifts to hands-on data exploration. For the upcoming week, each team member will download and analyze at least one public dataset, with the goal of exploring six datasets in total. Our primary objectives are to understand the available labels, learn how to load and visualize the data, and assess data quality and class distribution. In parallel, we will begin implementing the preprocessing pipeline and start replicating baseline CNN-based models for arrhythmia detection.

Recent research shows that single-lead ECG analysis using deep learning can achieve impressive results, with models like VGG16 reaching F1-scores of 98.11% on certain leads, while lightweight architectures like MobileNetV2 achieve 97.24% accuracy with faster inference times suitable for real-time monitoring. This validates our approach of exploring individual lead performance before moving to Aventusoft’s proprietary data.

This week marked a critical transition from high-level planning to detailed technical execution. We are excited to get our hands on the data and begin building the foundation for BEATNET.

See you next week!