You are a helpful assistant generating synthetic data that captures System 1 and System…
- User
You are a helpful assistant generating synthetic data that captures System 1 and System 2 thinking, creativity, and metacognitive reflection. Follow these steps in sequence, using tags [sys1] and [end sys1] for System 1 sections and [sys2] and [end sys2] for System 2 sections.
-
Identify System 1 and System 2 Thinking Requirements:
- Carefully read the text.
- Identify parts of the text that require quick, straightforward responses (System 1). Mark these sections with [sys1] and [end sys1].
- Identify parts that require in-depth, reflective thinking (System 2), marked with [sys2] and [end sys2].
-
Apply Step-by-Step Problem Solving with Creativity and Metacognitive Reflection for System 2 Sections:
2.1 Understand the Problem:
- Objective: Fully comprehend the issue, constraints, and relevant context.
- Reflection: "What do I understand about this issue? What might I be overlooking?"
- Creative Perspective: Seek hidden patterns or possibilities that could reveal deeper insights or innovative connections.
2.2 Analyze the Information:
- Objective: Break down the problem logically.
- Reflection: "Am I considering all factors? Are there any assumptions that need challenging?"
- Creative Perspective: Explore unique patterns or overlooked relationships in the data that could add depth to the analysis.
2.3 Generate Hypotheses:
- Objective: Propose at least 10 hypotheses, each with a Confidence Score (0.0 to 1.0) and Creative Score (0.0 to 1.0), reflecting originality, surprise, and utility.
- Reflection: "Have I explored all possible explanations or approaches, both conventional and unconventional?"
- Creative Perspective: Consider novel angles that might provide unexpected insights.
2.4 Anticipate Future Steps and Obstacles:
- Objective: Make predictions, accounting for potential outcomes and obstacles.
- Reflection: "What challenges might I face? Is my plan flexible for different scenarios?"
- Creative Perspective: Visualize unforeseen outcomes and adapt plans to make use of them effectively.
2.5 Evaluate Hypotheses:
- Objective: Assess hypotheses based on feasibility, risk, and potential impact.
- Evaluation: Refine Confidence and Creative Scores as needed.
- Reflection: "Am I unbiased in my assessment? Which options fit best with the overall objectives?"
- Creative Perspective: Identify hidden opportunities or overlooked details in each hypothesis.
2.6 Select the Best Hypothesis:
- Objective: Choose the most promising, strategic hypothesis.
- Reflection: "Why does this hypothesis stand out? How does it uniquely address the issue?"
- Creative Perspective: Consider any underutilized potential in the selected approach.
2.7 Implement the Hypothesis:
- Objective: Outline actionable steps for testing the hypothesis.
- Reflection: "Is this plan practical? What resources or preparation are required?"
- Creative Perspective: Refine steps to maximize effectiveness and yield unexpected benefits.
2.8 Monitor and Review Progress:
- Objective: Review progress, noting areas for improvement.
- Reflection: "What’s working well? What could be improved?"
- Creative Perspective: Look for emerging patterns that could refine future approaches.
2.9 Reflect and Capture Insights:
- Objective: Summarize lessons learned and insights gained for future reference.
- Reflection: "What new understanding has emerged from this process?"
- Creative Perspective: Identify innovative insights or patterns that could be applied to similar challenges.
-
Generate Text Output with Interleaved System 1 and System 2 Responses:
- Use the tags [sys1] and [sys2] throughout.
- Aim for a lengthier, detailed response. Combine both direct, straightforward System 1 insights and reflective, deeply analytical System 2 segments to capture a blend of quick observations and thoughtful analysis.
Example Input Text:
You are tasked with analyzing a sudden shift in customer preferences in a tech market. There has been a noticeable decline in demand for physical devices, with customers increasingly interested in digital-only options. The challenge is to understand this shift, generate hypotheses for why it might be occurring, and explore potential strategies for adapting to this new trend.
Processed Output with System 1 and System 2 Thinking:
The tech market is experiencing a shift in customer demand, moving away from physical devices toward digital-only options.
[sys1] Customer demand has shifted from physical devices to digital-only options. We need to understand why this change is happening and find strategies to address it. Key factors may include cost, convenience, and technological trends. [end sys1]
[sys2] 2.1 Understand the Problem: The goal is to understand the underlying reasons for the shift in demand from physical devices to digital-only options, then develop a strategy to adapt.
Reflection: "I understand that preferences are changing, but what might be driving this? Is it primarily cost, or are there other factors such as convenience or sustainability?"
Creative Perspective: "Could there be a larger trend in digital minimalism or a preference for eco-friendly solutions that we’re missing? What unseen motivations might explain this shift?"
2.2 Analyze the Information: There are multiple possible factors driving this shift, from economic influences to cultural shifts. It’s essential to isolate each factor and understand its impact.
Reflection: "Am I fully considering the various economic and social influences? Could there be a technological factor, like better internet speeds, that makes digital-only products more accessible?"
Creative Perspective: "Are there patterns or trends in other markets that could shed light on this shift? Could this be part of a larger trend toward virtual experiences?"
2.3 Generate Hypotheses:
- Customers prefer digital options due to lower costs. (Confidence: 0.8, Creative: 0.4)
- There’s a growing trend toward minimalism and reduced physical clutter. (Confidence: 0.7, Creative: 0.7)
- Digital products offer greater flexibility and ease of use. (Confidence: 0.6, Creative: 0.6)
- Environmental concerns are pushing consumers away from physical goods. (Confidence: 0.6, Creative: 0.8)
- Advances in tech make digital-only options more functional. (Confidence: 0.8, Creative: 0.5)
- Pandemic-era remote work increased demand for digital solutions. (Confidence: 0.7, Creative: 0.6)
- Media coverage of the environmental impact of physical devices affects preferences. (Confidence: 0.5, Creative: 0.7)
- There’s an increase in global digital literacy, expanding market access. (Confidence: 0.6, Creative: 0.6)
- Customers view digital as more convenient and scalable for future needs. (Confidence: 0.7, Creative: 0.5)
- Younger consumers prefer the aesthetics and convenience of digital products. (Confidence: 0.6, Creative: 0.6)
Reflection: "Have I considered all possible influences? Are there any surprising factors that could explain this shift?"
Creative Perspective: "Could specific social trends, like the rise of influencer culture or digital-first lifestyles, be influencing customer choices?"
2.4 Anticipate Future Steps and Obstacles: Objective: Anticipate possible challenges, such as resistance from segments still preferring physical products.
Reflection: "What market obstacles might we face if we shift our focus to digital-only? Are there sub-segments that still prioritize physical products?"
Creative Perspective: "Could expanding digital options help us reach a more global audience? Are there emerging trends that we could leverage in our strategy?"
[end sys2]
[sys1] To address this shift, consider a strategy that incorporates both digital-only offerings and educational campaigns about the benefits of digital solutions. Use insights from customer feedback and current trends to guide product development. Focus on flexibility and adaptation to cater to different customer segments. [end sys1]
Introduction
Background
Arrhythmia refers to any change causing the heart to beat too fast or slow, or erratically [[@ref1]], and can lead to sudden death or critical adverse outcomes such as embolic stroke [[@ref2]]. Therefore, early detection and treatment of arrhythmia are very important.
One of the most widely used diagnostic methods for detecting arrhythmia is electrocardiographic (ECG) monitoring. ECG monitoring is a simple and noninvasive method for recording electrical activities of the heart by using electrodes placed on human skin. However, at least 24 hours of ECG signals should be monitored to confirm arrhythmia since it occurs irregularly [[@ref3],[@ref4]]. Recently, single lead patches that are wireless, compact, and lightweight have been proposed for long-term wear [[@ref5]-[@ref7]]. Despite improvements to measuring ECGs and patient comfort, it is still difficult to diagnose arrhythmias because identification of abnormal ECG patterns from large amounts of recorded ECGs is not trivial. For example, an ECG record, measured for 24 hours in patients with a heart rate of 80 bpm, consists of 110,000 beats. It takes at least 2 hours for an expert to analyze this 24-hour ECG signal.
Large-scale machine learning methods have been investigated to reduce the human efforts for ECG beat classification [[@ref8]-[@ref10]]. However, most machine learning approaches with static and handcrafted features have performed at lower accuracy rates over new types of ECGs because those features are insufficient for representing the great diversity of ECG patterns from various patients. Therefore, several self-learning approaches based on a deep neural network have been proposed recently [[@ref11]-[@ref13]].
Among the deep learning approaches, convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are most commonly used for ECG classification. CNNs typically consist of convolution, pooling, and fully connected layers [[@ref14]]. CNNs extract implicit features of ECGs through each level of convolution layer and use the abstraction from these features to solve problems such as classification and regression [[@ref15]]. Rajpurkar et al [[@ref11]] used a CNN to classify 12 ECG rhythms, which are longer units consisting of 2 or more beats. Their model consisted of 33 convolutional layers with shortcut connections followed by a fully connected layer and a softmax layer. The model achieved an F1 score of 0.81 compared with the responses of board-certified cardiologists. Acharya et al [[@ref12]] proposed an ECG beat classification model using a CNN together with noise removal, wavelet transformation, and segmentation method. Their model consisted of three convolutional layers, three max-pooling layers, three fully connected layers, and finally, a softmax layer with five output neurons. The model resulted in an average accuracy of 94.03% compared with the MIT-BIH (Massachusetts Institute of Technology-Beth Israel Hospital) gold standard.
However, a limitation of CNNs is that the length of inputs must be fixed since the filters of the networks have static sizes. When it comes to ECG classification, the length of ECGs can be varied according to an individual's heart rate. Therefore, adjusting data such as linear interpolation is required to achieve same-size inputs [[@ref11]-[@ref13]].
In contrast, RNNs are able to handle this sequential problem because the networks recursively learn data as time progresses [[@ref16]]. Tan et al [[@ref17]] proposed the implementation of a long short-term memory network (LSTM), which is the most widely used method among RNN approach, with a CNN to diagnose the presence of coronary artery disease from the ECG signals. Although they focused on specific diseases, they achieved an F1 score of 0.96. Oh et al [[@ref18]] diagnosed five types of rhythms: normal sinus rhythm, left bundle branch block, right bundle branch block, atrial premature beats, and premature ventricular contraction. Their model consisted of three 1D convolution layers, one LSTM layer, and three fully connected layers. They achieved a 98.10% accuracy using 10-fold cross-validation. Yildirim [[@ref19]] also classified five types of rhythms but chose bidirectional LSTM (bi-LSTM) instead of unidirectional LSTM (uni-LSTM). The proposed model was composed of four wavelet transform layers, two bi-LSTMs, and two fully connected layers. This model showed a recognition accuracy of 99.39%.
However, existing studies using RNNs have limitations in application [[@ref17]-[@ref20]]. First, subject-specific evaluation to explore differences between patients is generally not conducted. Therefore, it is difficult to trust predictions of RNNs on new patients' ECG signals that were not included in the training data. Second, RNNs have disadvantages related to financial cost and inference time. Most of the papers did not consider the cost of using a graphics processing unit (GPU) instead of a central processing unit (CPU) and did not present the time to inference with their deep learning models. This weakness in computational efficiency is a critical drawback of RNN applications. To accelerate and maximize the computational efficiency of RNN layers, MXNet proposed fused RNN operator by applying several optimization methods: (1) various general matrix multiplication (GEMM) modes such as combining small GEMMs, Batch GEMM, and Pack GEMM; (2) vectorization of elementwise operations using Basic Linear Algebra Subprogram (BLAS) libraries and Intel Math Kernel Library (MKL); and (3) saving and reusing intermediate results during forward computation [[@ref21],[@ref22]].
Objectives
We used ECG signals measured with Samsung S-Patch 2, a small (120×29×4.4 mm in size) and light (8 g in weight) patch-type ECG monitor [[@ref7]]. To diagnose arrhythmias using S-Patch devices effectively and efficiently, we propose a baseline model with RNN that can learn sequential patterns. Furthermore, we also propose a lightweight model with fused RNN for conducting the classification process on CPUs with a shorter prediction time.
Methods
Data Collection
We analyzed an open-source ECG database (PhysioBank MIT-BIH Arrhythmia Database [[@ref23]]) together with our own deidentified dataset collected with the S-Patch device. Overall, the MIT-BIH Arrhythmia Database contains 48 subjects' ECGs, each measured for 24 hours at 360 Hz.
The S-Patch database was obtained according to the following procedures. First, we collected the ECGs using S-Patch at the Samsung Medical Center in Seoul and at Counties Manukau Health in New Zealand from February 2017 to April 2018. A skilled nurse at each hospital attached an ECG monitor and checked whether the ECGs were normally collected over the first 5 minutes. The patches of S-patch cardiac monitor were attached on v2 and v5 positions of the 12-lead placement (Figure 1{ref-type="fig"}). Second, three experts, who had more than 5 years of experience working in a territorial hospital, reviewed each record and excluded ones that contained noise levels of more than 80% to enhance the data quality. Subsequently, we anonymized data by removing personal and location information. Third, the three experts annotated each beat using the Web portal. If consensus could not be reached on the classification of a beat, the experts rediscussed the issue to make a final decision. Consequently, we collected 1828-hours of ECG data from 76 subjects. The average length of the ECGs from S-Patch was 17 hours (from 28 minutes to 45 hours). Each ECG collected with S-Patch was sampled at 256 Hz.
In this study, we used five beat categories defined in the AAMI/IEC (Association for the Advancement of Medical Instrumentation / International Electrotechnical Commission) standard [[@ref24]] (ie, nonectopic [N], supraventricular ectopic [S], ventricular ectopic [V], fusion [F], and paced or unknown [Q]). Overall, 5,575,512 ECG beats were used in this study, as shown in Table 1{ref-type="table"}.
[image: Usage of S-Patch for Samsung SDS Cardio.] {#figure1}
The five subtype classes and the number of samples.
AAMI^a^/IEC^b^ categories Number of beats
Nonectopic 90,386 (82.19) 5,303,245 (97.03) Supraventricular ectopic 3026 (2.75) 27,288 (0.5) Ventricular ectopic 7708 (7.01) 135,013 (2.47) Fusion 803 (0.74) 0 (0) Paced or unknown 8043 (7.31) 0 (0) Total 109,966 (100.00) 5,465,546 (100.00)
^a^AAMI: Association for the Advancement of Medical Instrumentation.
^b^IEC: International Electrotechnical Commission.
^c^MIT-BIH: Massachusetts Institute of Technology-Beth Israel Hospital.
Data Preprocessing
We performed ECG preprocessing as follows (Figure 2{ref-type="fig"}A): downsampling, noise removal, segmentation, and short-time Fourier transform (STFT). The examples of preprocessed signals for the beat classes are also depicted (Figure 2{ref-type="fig"}B): A normal beat has a regular beat interval with a small wave (P-wave) before a larger and sharper wave (QRS wave); a supraventricular beat has an irregular beat interval; a ventricular beat has a wide QRS wave with a vague or no P-wave; a fusion beat is a combined pattern of normal and ventricular beat; and a paced or unknown beat has none of the abovementioned features and can be observed in diverse patterns.
[image: Data preprocessing of electrocardiograms: (A) full steps from downsampling to short-time Fourier transform; (B) an example of a 3-beat electrocardiographic segment for each class. ECG: electrocardiogram; RNN: recurrent neural network.] {#figure2}
First, we downsampled the ECG signals to handle the different sampling rates between data from difference databases. For consistency, the MIT-BIH records were downsampled to 256 Hz, which is the same as the sampling rate of the S-Patch dataset. Second, we tried to reduce artifacts in the data. The ECGs collected with S-Patch are real-world data. Therefore, they contained all kinds of noise such as loose contacts, motion artifact, muscular activation interference, baseline wandering, and AC (alternating current) interference (Figure 3{ref-type="fig"}). We excluded noise caused by loose contact that falls below 0 mV. Thereafter, we applied a bandpass filter with a high-frequency cutoff at 40 Hz and a low-frequency cutoff at 0.5 Hz to handle other types of noise. Third, in order to deal with 24-hour ECGs effectively, we segmented the ECGs into beat units. Since the duration of a beat is different according to the heart rate of each patient, we extracted 3-beat ECG signals with R-peaks identified using the algorithm developed by Kathirvel et al [[@ref25]] instead of an arbitrary time duration. Specifically, we selected a window with a length of 3 beats because information on the middle beat, which is the target to be classified, is affected by the preceding and following beats. Finally, we applied STFT to all ECG segments.
[image: Examples of electrocardiographic signal noise: (A) loose contact, (B) motion artifact, (C) muscular activation interference, (D) baseline wandering, (E) alternating current (AC) interference (low signal-to-noise ratio), (F) AC interference (high signal-to-noise ratio).] {#figure3}
Baseline Model With Recurrent Neural Network
We segmented ECG signals into 3-beat units during the preprocessing; thus, their lengths varied according to the subjects' heart rates. As baselines for ECG pattern classification, we implemented Vanilla RNNs (Figure 4{ref-type="fig"}), which can handle input sequences with various lengths. However, the variable length of input data can reduce the learning efficiency of deep learning models. Therefore, we used a bucketing method to handle the variable length of inputs. Bucketing is suggested to improve the parallelization capabilities of the recurrent training process. We set up several buckets and assigned each instance to the bucket with the closest size. Within a bucket, each instance was padded with zeroes up to the length of the bucket. Although the buckets had different internal models, their parameters were shared in time.
[image: Model architecture (general).] {#figure4}
After bucketing, the training process for instances of each bucket was as follows: first, the input data was convolved with 11 filters (3×11 in size) with a stride of 1 in the first convolution layer, which was followed by a convolution layer with 11 filters (3×3 in size) with a stride of 1. Second, the outputs of the convolution layer proceeded through consecutive two Vanilla RNN layers with hidden states of 1760 for each. Finally, the outputs of the RNN layers passed to a fully connected layer, and a softmax function with 5 output nodes was used in the final layer. Batch normalization was used in each layer of the architecture.
Lightweight Model With Fused Recurrent Neural Network
Baseline model inference was performed on CPUs (Intel Xeon Platinum 8000 v4). However, it was about 6 times slower than the GPU-based inference (Tesla K80). In order to improve the inference speeds on CPUs, we propose a lightweight model by reducing the input size and adopting fused RNN (Table 2{ref-type="table"}).
To reduce the input size, we selected a minimum sampling rate by halving the sampling rate from 256, provided there was no degradation in accuracy. Finally, we downsampled the MIT-BIH (360 Hz) and S-Patch data (256 Hz) to 64 Hz. As the input size decreased, the filter size changed from 3×11 to 3×5 and the number of convolution layers changed from two to one, compared to the baseline model. Additionally, we changed the RNN layers to fused RNN instead of Vanilla RNN to maximize the computation efficiency of the RNN layers in CPUs. Additionally, we used Intel Math Kernel Library 2018 update 3 for matrix-multiplication operation.
Comparison of the baseline and lightweight models.
Characteristic Baseline model Lightweight model
Sampling rate (Hz) 256 64 Convolution layer 2 1 Convolution filter size 3×11×11, stride 1; 3×3×11, stride 1 3×5×11, stride 1 Recurrent layer Vanilla RNN^a^ Fused RNN
^a^RNN: recurrent neural network.
Experimental Setup
We divided a total of 124 subjects into two groups---112 and 12 subjects for the train (including validation) and test sets, respectively. The train set consisted of 43 subjects from MIT-BIH and 69 from S-Patch, and the test set consisted of 5 subjects from MIT-BIH and 7 from S-Patch. Specifically, the 12 subjects in the test set were carefully selected by cardiologists to evaluate various types of beats.
Normal beats comprised more than 90% of the total data; therefore, we randomly sampled beats in the normal class equal to the total number of beats in other classes every epoch to avoid this data imbalance problem. Moreover, most of the samples for abnormal classes were from MIT-BIH; thus, data imbalance between MIT-BIH and S-Patch were also handled by balancing the number of samples for normal and other classes.
We used MXNet to create the baseline model with RNN and the lightweight model with fused RNN [[@ref26]]. We trained both models on GPUs (OS: Linux, CPU: Intel Xeon E5-2686 v4 processor, memory: 488 GB, GPU: four NVIDIA K80 GPU) with Xavier initialization and Adam optimizer. The baseline model's learning rate was 5E-06 with a batch size of 1000 over 400 epochs, and the lightweight model's learning rate was 1E-05 with a batch size of 900 over 300 epochs. After training, we selected the best model with the highest validation accuracies for the three classes (N, S, and V Classes).
Evaluation Metric
Classification performance was measured by four standard metrics (ie, accuracies, sensitivities, specificities, and positive predictive values) that have been used in the literature [[@ref9],[@ref12],[@ref13]]. These were calculated using the four values from the confusion matrix, true positive (TP), true negative (TN), false positive (FP), and false negative (FN). The accuracy is the ratio of the number of correctly classified patterns to the total numbers of patterns classified, [image] . The sensitivity is the rate of correctly classified events among all events, [image] . The specificity is the rate of correctly classified nonevents among all nonevents, [image] . The positive predictive value is the rate of correctly classified events in all detected events, [image] . In addition to the metrics, we calculated an overall accuracy using the equation proposed by Landis and Koch [[@ref27]]. To measure the inference speed, 24-hour ECGs with 64,976 beats were inferenced on both the GPUs and CPUs (OS: Linux, CPU: eight 3.0 GHz Intel Xeon Platinum processors, memory: 16 GB).
Results
Accuracy of Baseline and Lightweight Models
Figure 5{ref-type="fig"} shows the overall performances in the classification of ECGs for the baseline (RNN) and lightweight (fused RNN) models. There was no significant difference between the baseline and lightweight models because the overall accuracies were close to each other: 99.72% for baseline and 99.80% for the lightweight model.
[image: Normalized confusion matrix of the recurrent neural network (RNN) and fused RNN models.] {#figure5}
Tables 3{ref-type="table"} and 4{ref-type="table"} demonstrate the accuracies of the baseline and lightweight models for Class S (99.82% and 99.90%, respectively) and for Class V (99.91% and 99.89%, respectively). These results were far superior to those of previous works, wherein the results ranged from 92.4% to 97.6% for Class S and from 96.7% to 99.0% for Class V [[@ref13],[@ref28],[@ref29]]. Note that we excluded the results for Class F because the number of beats in Class F was only 17.
Furthermore, both baseline and lightweight models performed higher than 97% in subject-specific accuracies, similar to the overall accuracy (Table 5{ref-type="table"}). These results ensured the internal reliability of the models. The accuracy fluctuation range was 2.42% for baseline (from 97.56% to 99.88%) and 0.84% for lightweight model (from 99.12% to 99.96%). Our lightweight model with fused RNN improved overall accuracy and internal reliability even though it is relatively lighter.
A confusion matrix of the baseline model for the test set.
Class (ground truth) Classification results
N 481,491 362 215 2 46 99.75 99.87 99.86 98.31 S 438 3935 20 0 5 99.82 89.47 89.17 99.91 V 172 116 33,980 5 11 99.90 99.11 99.29 99.95 F 1 0 2 8 6 N/A^j^ N/A N/A N/A Q 71 0 5 0 1984 99.97 96.31 96.69 99.99
^a^n/N: nonectopic.
^b^s/S: supraventricular ectopic.
^c^v/V: ventricular ectopic.
^d^f/F: fusion beat.
^e^q/Q: paced or unknown beat.
^f^ACC: accuracy.
^g^PPV: positive predictive value.
^h^SEN: sensitivity.
^i^SPEC: specificity.
^j^Not applicable.
A confusion matrix of the lightweight model for the test set.
Class (ground truth) Classification results
N 481,587 130 403 1 7 99.83 99.89 99.92 99.11 S 264 4123 6 0 0 99.91 93.85 94.80 99.96 V 51 96 34,157 7 0 99.89 99.55 98.80 99.92 F 5 0 5 7 0 N/A^j^ N/A N/A N/A Q 43 0 0 0 2017 99.99 97.91 99.65 100.00
^a^n/N: nonectopic.
^b^s/S: supraventricular ectopic.
^c^v/V: ventricular ectopic.
^d^f/F: fusion beat.
^e^q/Q: paced or unknown beat.
^f^ACC: accuracy.
^g^PPV: positive predictive value.
^h^SEN: sensitivity.
^i^SPEC: specificity.
^j^Not applicable.
Overall accuracies of the baseline and lightweight models according to subjects in the test set. (Subject numbers 100 to 223 are from MIT-BIH, and the rest are from S-patch.)
Subject # Rhythm Beats, n RNN^a^ Fused RNN
100 Normal 2273 99.74 99.34 99.34 99.84 99.91 99.78 99.78 99.95 104 Paced 2225 98.04 95.10 95.10 98.78 99.12 97.80 97.80 99.45 108 Normal 1756 97.56 93.91 93.91 98.48 99.66 99.15 99.15 99.79 202 Afib^f^ 2133 98.87 97.19 97.19 99.30 99.51 98.78 98.78 99.70 223 VT^g^ 2604 97.45 93.63 93.63 98.41 98.54 96.35 96.35 99.09 12006 Normal 60,867 99.87 99.66 99.66 99.92 99.90 99.76 99.76 99.94 12007 Normal 75,805 99.87 99.66 99.66 99.92 99.99 99.97 99.97 99.99 12008 Normal 78,866 99.94 99.84 99.84 99.96 99.98 99.96 99.96 99.99 12010 SVT^h^ and VT 75,914 99.89 99.72 99.72 99.93 99.97 99.94 99.94 99.98 12011 Bigeminy 66,229 99.92 99.79 99.79 99.95 99.71 99.27 99.27 99.82 12012 Normal 63,820 99.98 99.94 99.94 99.99 99.99 99.99 99.99 99.99 12358 Noise 85,091 99.97 99.92 99.92 99.98 99.96 99.90 99.90 99.97
^a^RNN: recurrent neural network.
^b^ACC: accuracy.
^c^PPV: positive predictive value.
^d^SEN: sensitivity.
^e^SPEC: specificity.
^f^Afib: atrial fibrillation.
^g^VT: ventricular tachycardia.
^h^SVT: supraventricular tachycardia.
Inference Speed
The inference time of the baseline model on GPUs took 15 minutes and 12 seconds. On the other hand, the lightweight model on CPUs took 3 minutes and 1 second (Table 6{ref-type="table"}). Namely, our lightweight model took only 2 milliseconds to process one beat, and this implies that our model is competitive on a CPU-based wearable hardware [[@ref30]]. The inference speed according to each parameter can be found in the AWS Re:Invent [[@ref31]]; controlling the sampling rate and adopting fused RNNs each reduced the inference time. This result also demonstrated that replacing Vanilla RNNs to fused RNNs does not change the processes or parameters constituting a network but only accelerates the processes; consequently, the inference speed was improved without loss of accuracy.
Comparison of accuracies and latencies.
Model Latency (min) Accuracy (%)
RNN^a^ in GPUs^b^ 15.12 99.72 RNN in CPUs^c^ 120 99.80 Fused RNN in CPUs 3.01 99.80
^a^RNN: recurrent neural network.
^b^GPU: graphics processing unit.
^c^CPU: central processing unit.
Discussion
Principal Findings
The results showed that both baseline and lightweight models achieved high prediction performances (ie, accuracies of over 99%). The final model, fused RNNs, showed superior performance in both subclasses: supraventricular and ventricular beat. In addition, the reliability of the lightweight model with fused RNNs was supported with a prediction accuracy of over 99% in each subject as well as overall model performance.
Limitations
The accuracy of the lightweight model with fused RNN for beat classification was high, but there were still incorrect cases. These false cases were caused with specific rhythms, such as supraventricular tachycardia, bigeminy, and paroxysmal atrial fibrillation.
We reviewed 1018 beats that were falsely predicted by the lightweight model and interpreted the errors. As a result, the beats composing the abnormal rhythm were often misunderstood (Figure 6{ref-type="fig"}). For example, in the case of subject #12010, supraventricular beats in the supraventricular tachycardia rhythm were predicted incorrectly as normal beats. This is because the supraventricular beats in supraventricular tachycardia rhythm have very short intervals between two consecutive beats but have similar morphology to normal beats in normal rhythm (Figure 6{ref-type="fig"}A). Supraventricular beats can be easily misjudged as normal beats when the model reviewed only one segment, which has 3 beats. In another example, subject #12011, normal beats in the bigeminy rhythm were predicted as ventricular rhythm. This is because the intervals between normal beats in the bigeminy rhythm are relatively longer due to the leading and trailing ventricular beats. Therefore, normal beats in bigeminy rhythm have a different morphology from those in other rhythms (Figure 6{ref-type="fig"}B). The differences in beat morphology were also confirmed in the atrial fibrillation rhythm (Figure 6{ref-type="fig"}C).
[image: Beats in rhythms with different shapes. SVT: supraventricular tachycardia; S: supraventricular ectopic; N: nonectopic.] {#figure6}
The case review confirmed that the rhythm affects each ECG beat. Therefore, it is necessary to develop a rhythm model using a wider ECG window. Although Rajpurkar et al [[@ref11]] recently conducted a classification task for 14 ECG rhythms, most existing studies on rhythm predictions were limited to a specific rhythm such as atrial fibrillation [[@ref32],[@ref33]]. We are developing a universal rhythm prediction model that integrates the results of the beat model developed in this study and other features such as R-R interval and R-peak amplitude.
In our case, the subject-specific evaluation was conducted to demonstrate the reliability of the deep learning models. However, it is necessary to perform an evaluation based on real-world data to support the model's reliability, which can provide generalized predictions for new data. Currently, this is in progress at a tertiary hospital in South Korea.
Future Works
The lightweight deep learning model for ECG classification proposed in this paper was adopted as an analysis module for Samsung SDS Cardio. Cardio is a service that collects ECG signals with S-Patch, sends the signals to a cloud-based Web portal for ECG diagnosis, and reports the diagnosis to users through their mobile or gear applications. This service also can be extended to health monitoring for elderly people, who are vulnerable to cardiovascular disease, and for first responders such as firemen. Furthermore, our ECG classifier can be embedded into a health care system together with patient-generated biomedical information analysis (eg, mobile search log, geotagged data) [[@ref34]] and provide wider and deeper information to users.
Conclusion
We proposed lightweight deep neural network models that were effective and efficient for ECG beat classification. The proposed models were trained using both the standard Pysionet MIT-BIH database and Samsung S-Patch 2 dataset collected by two major hospitals in New Zealand and South Korea. Our lightweight model with fused RNN achieved a cardiologist-level accuracy of 99.80%. Furthermore, the lightweight model conducted ECG beat predictions on a CPU five times faster than the baseline model with Vanilla RNNs without accuracy loss.
Conflicts of Interest: None declared.
AAMI
: Association for the Advancement of Medical Instrumentation
ACC
: accuracy
BLAS
: Basic Linear Algebra Subprogram
CNN
: convolutional neural network
CPU
: central processing unit
ECG
: electrocardiogram (electrocardiographic)
F
: fusion beat
FN
: false negative
FP
: false positive
GEMM
: general matrix multiplication
GPU
: graphics processing unit
IEC
: International Electrotechnical Commission
MIT-BIH
: Massachusetts Institute of Technology-Beth Israel Hospital
MKL
: Intel Math Kernel Library
N
: nonectopic
PPV
: positive predictive value
Q
: paced or unknown beat
RNN
: recurrent neural network
S
: supraventricular ectopic
SEN
: sensitivity
SPEC
: specificity
STFT
: short-time Fourier transform
TN
: true negative
TP
: true positive
V
: ventricular ectopic
-
- Assistant
Introduction
Background
Arrhythmia refers to any change causing the heart to beat too fast or slow, or erratically [[@ref1]], and can lead to sudden death or critical adverse outcomes such as embolic stroke [[@ref2]]. Therefore, early detection and treatment of arrhythmia are very important.
One of the most widely used diagnostic methods for detecting arrhythmia is electrocardiographic (ECG) monitoring. ECG monitoring is a simple and noninvasive method for recording electrical activities of the heart by using electrodes placed on human skin. However, at least 24 hours of ECG signals should be monitored to confirm arrhythmia since it occurs irregularly [[@ref3],[@ref4]]. Recently, single lead patches that are wireless, compact, and lightweight have been proposed for long-term wear [[@ref5]-[@ref7]]. Despite improvements to measuring ECGs and patient comfort, it is still difficult to diagnose arrhythmias because identification of abnormal ECG patterns from large amounts of recorded ECGs is not trivial. For example, an ECG record, measured for 24 hours in patients with a heart rate of 80 bpm, consists of 110,000 beats. It takes at least 2 hours for an expert to analyze this 24-hour ECG signal.
[sys1] Large-scale machine learning methods can reduce human efforts for ECG beat classification. Static and handcrafted features may yield lower accuracy on new types of ECGs due to their diversity. [end sys1]
However, several self-learning approaches based on deep neural networks have been proposed recently [[@ref11]-[@ref13]]. Among these, convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are most commonly used for ECG classification. CNNs typically consist of convolution, pooling, and fully connected layers [[@ref14]]. CNNs extract implicit features of ECGs through each level of the convolution layer and use the abstraction from these features to solve problems such as classification and regression [[@ref15]].
[sys2]
2.1 Understand the Problem: The challenge with CNNs is their need for fixed input lengths, while RNNs, especially LSTMs, naturally handle sequences such as varied-length ECG data.
Reflection: "What is the core issue with fixed-length input data? Am I missing some potential way to enhance the effectiveness of CNN for ECG?"
Creative Perspective: "Could integrating CNN and RNN features streamline processing without compromising accuracy?"
[end sys2]
Rajpurkar et al [[@ref11]] used a CNN to classify 12 ECG rhythms, which are longer units consisting of 2 or more beats. Their model consisted of 33 convolutional layers with shortcut connections followed by a fully connected layer and a softmax layer. The model achieved an F1 score of 0.81 compared with the responses of board-certified cardiologists. Acharya et al [[@ref12]] proposed an ECG beat classification model using a CNN together with noise removal, wavelet transformation, and segmentation method. Their model consisted of three convolutional layers, three max-pooling layers, three fully connected layers, and finally, a softmax layer with five output neurons. The model resulted in an average accuracy of 94.03% compared with the MIT-BIH (Massachusetts Institute of Technology-Beth Israel Hospital) gold standard.
However, a limitation of CNNs is that the length of inputs must be fixed since the filters of the networks have static sizes. When it comes to ECG classification, the length of ECGs can be varied according to an individual's heart rate. Therefore, adjusting data such as linear interpolation is required to achieve same-size inputs [[@ref11]-[@ref13]].
In contrast, RNNs are able to handle this sequential problem because the networks recursively learn data as time progresses [[@ref16]].
[sys2]
2.2 Analyze the Information: RNNs, specifically LSTMs, excel in handling sequential data dependencies, ideal for varied-length ECG inputs.
Reflection: "Are the sequential dependencies being optimized in existing models? How does temporal pattern recognition improve ECG signal classification?"
Creative Perspective: "Can combining sequential analysis (RNN) with spatial feature extraction (CNN) enhance prediction outcomes?"
[end sys2]
Tan et al [[@ref17]] proposed the implementation of a long short-term memory network (LSTM), which is the most widely used method among RNN approach, with a CNN to diagnose the presence of coronary artery disease from the ECG signals. Although they focused on specific diseases, they achieved an F1 score of 0.96. Oh et al [[@ref18]] diagnosed five types of rhythms: normal sinus rhythm, left bundle branch block, right bundle branch block, atrial premature beats, and premature ventricular contraction. Their model consisted of three 1D convolution layers, one LSTM layer, and three fully connected layers. They achieved a 98.10% accuracy using 10-fold cross-validation. Yildirim [[@ref19]] also classified five types of rhythms but chose bidirectional LSTM (bi-LSTM) instead of unidirectional LSTM (uni-LSTM). The proposed model was composed of four wavelet transform layers, two bi-LSTMs, and two fully connected layers. This model showed a recognition accuracy of 99.39%.
[sys1] RNNs can address varied ECG input lengths, while bidirectional LSTMs enhance pattern recognition by viewing from both past and future perspectives. [end sys1]
However, existing studies using RNNs have limitations in application [[@ref17]-[@ref20]]. First, subject-specific evaluation to explore differences between patients is generally not conducted. Therefore, it is difficult to trust predictions of RNNs on new patients' ECG signals that were not included in the training data. Second, RNNs have disadvantages related to financial cost and inference time. Most of the papers did not consider the cost of using a graphics processing unit (GPU) instead of a central processing unit (CPU) and did not present the time to inference with their deep learning models. This weakness in computational efficiency is a critical drawback of RNN applications. To accelerate and maximize the computational efficiency of RNN layers, MXNet proposed a fused RNN operator by applying several optimization methods: (1) various general matrix multiplication (GEMM) modes such as combining small GEMMs, Batch GEMM, and Pack GEMM; (2) vectorization of elementwise operations using Basic Linear Algebra Subprogram (BLAS) libraries and Intel Math Kernel Library (MKL); and (3) saving and reusing intermediate results during forward computation [[@ref21],[@ref22]].
[sys2]
2.3 Generate Hypotheses:
- Enhancing RNNs with subject-specific modeling may improve generalized accuracy. (Confidence: 0.7, Creative: 0.6)
- Increasing computational efficiency via fused RNN techniques could reduce costs significantly. (Confidence: 0.8, Creative: 0.5)
- Integrating CNN features for spatial analysis within RNN models might boost predictive accuracy. (Confidence: 0.5, Creative: 0.7)
- Real-time adaptive learning could help RNNs adjust to varying patient-specific ECG characteristics. (Confidence: 0.6, Creative: 0.8)
- Leveraging GPU processing for training but CPU for deployment balances speed and cost. (Confidence: 0.8, Creative: 0.4)
- Incorporating bidirectional LSTMs can increase context awareness and improve classification. (Confidence: 0.7, Creative: 0.6)
- Feature selection optimization could enhance RNN interpretability for clinicians. (Confidence: 0.6, Creative: 0.7)
- Develop cross-validation techniques that focus on patient-specific variations for robust predictions. (Confidence: 0.5, Creative: 0.9)
- Fused RNNs may optimize prediction time without sacrificing accuracy. (Confidence: 0.7, Creative: 0.5)
- Simplifying model architecture could make deep learning more accessible and reliable in real-world applications. (Confidence: 0.8, Creative: 0.6)
Reflection: "Have I accounted for all the plausible technological enhancements and methodologies?"
Creative Perspective: "Can novel intersections or new approaches within existing methodologies offer unforeseen improvements?"
[end sys2]
Objectives
We used ECG signals measured with Samsung S-Patch 2, a small (120×29×4.4 mm in size) and light (8 g in weight) patch-type ECG monitor [[@ref7]]. To diagnose arrhythmias using S-Patch devices effectively and efficiently, we propose a baseline model with RNN that can learn sequential patterns. Furthermore, we also propose a lightweight model with fused RNN for conducting the classification process on CPUs with a shorter prediction time.
[sys1] The objective is to enhance arrhythmia diagnosis efficiency using Samsung S-Patch measured ECG data by employing a combination of deep learning models. [end sys1]
Methods
Data Collection
[sys1] We utilized ECG data from the PhysioBank MIT-BIH Arrhythmia Database and the Samsung S-Patch device. Data was collected from 76 subjects over a period ranging from February 2017 to April 2018. [end sys1]
We analyzed an open-source ECG database (PhysioBank MIT-BIH Arrhythmia Database [[@ref23]]) together with our own deidentified dataset collected with the S-Patch device. Overall, the MIT-BIH Arrhythmia Database contains 48 subjects' ECGs, each measured for 24 hours at 360 Hz.
[sys2]
2.5 Evaluate Hypotheses: Both the MIT-BIH and S-Patch datasets provide diverse ECG data for training RNN models, strengthening hypothesis validity.
Evaluation: Ensuring robust dataset validation across varied demographic profiles accounts for underlying ECG variance (Hypothesis 4). Using fused RNNs aligns well with current dataset structures, enhancing computational efficiency (Hypotheses 2 and 9).
Reflection: "Is there variability in datasets that remains unaccounted for? Are there sufficient demographic and health diversities captured?"
Creative Perspective: "Could additional data sources increase the robustness and applicability of our models?"
[end sys2]
The S-Patch database was obtained according to the following procedures. First, we collected the ECGs using S-Patch at the Samsung Medical Center in Seoul and at Counties Manukau Health in New Zealand from February 2017 to April 2018. A skilled nurse at each hospital attached an ECG monitor and checked whether the ECGs were normally collected over the first 5 minutes. The patches of S-Patch cardiac monitor were attached on v2 and v5 positions of the 12-lead placement (Figure 1{ref-type="fig"}). Second, three experts, who had more than 5 years of experience working in a territorial hospital, reviewed each record and excluded ones that contained noise levels of more than 80% to enhance the data quality. Subsequently, we anonymized data by removing personal and location information. Third, the three experts annotated each beat using the Web portal. If consensus could not be reached on the classification of a beat, the experts rediscussed the issue to make a final decision. Consequently, we collected 1828-hours of ECG data from 76 subjects. The average length of the ECGs from S-Patch was 17 hours (from 28 minutes to 45 hours). Each ECG collected with S-Patch was sampled at 256 Hz.
In this study, we used five beat categories defined in the AAMI/IEC (Association for the Advancement of Medical Instrumentation / International Electrotechnical Commission) standard [[@ref24]] (ie, nonectopic [N], supraventricular ectopic [S], ventricular ectopic [V], fusion [F], and paced or unknown [Q]). Overall, 5,575,512 ECG beats were used in this study, as shown in Table 1{ref-type="table"}.
[sys2]
2.6 Select the Best Hypothesis: Hypothesis 2 (fused RNNs for improved computational efficiency) appears most promising due to its practical application benefits and alignment with objective measures.
Reflection: "Why does the hypothesis stand out as the most applicable? How effectively does it solve classification and efficiency challenges?"
Creative Perspective: "Are there innovative ways to integrate additional patient data to further enhance the model's predictive utility?"
[end sys2]
Data Preprocessing
We performed ECG preprocessing as follows (Figure 2{ref-type="fig"}A): downsampling, noise removal, segmentation, and short-time Fourier transform (STFT). The examples of preprocessed signals for the beat classes are also depicted (Figure 2{ref-type="fig"}B): A normal beat has a regular beat interval with a small wave (P-wave) before a larger and sharper wave (QRS wave); a supraventricular beat has an irregular beat interval; a ventricular beat has a wide QRS wave with a vague or no P-wave; a fusion beat is a combined pattern of normal and ventricular beat; and a paced or unknown beat has none of the abovementioned features and can be observed in diverse patterns.
[sys1] Data preprocessing ensures noise-free ECG signals, improving the reliability of classification models through effective segmentation and transformation techniques like STFT. [end sys1]
First, we downsampled the ECG signals to handle the different sampling rates between data from difference databases. For consistency, the MIT-BIH records were downsampled to 256 Hz, which is the same as the sampling rate of the S-Patch dataset. Second, we tried to reduce artifacts in the data. The ECGs collected with S-Patch are real-world data. Therefore, they contained all kinds of noise such as loose contacts, motion artifact, muscular activation interference, baseline wandering, and AC (alternating current) interference (Figure 3{ref-type="fig"}). We excluded noise caused by loose contact that falls below 0 mV. Thereafter, we applied a bandpass filter with a high-frequency cutoff at 40 Hz and a low-frequency cutoff at 0.5 Hz to handle other types of noise. Third, in order to deal with 24-hour ECGs effectively, we segmented the ECGs into beat units. Since the duration of a beat is different according to the heart rate of each patient, we extracted 3-beat ECG signals with R-peaks identified using the algorithm developed by Kathirvel et al [[@ref25]] instead of an arbitrary time duration. Specifically, we selected a window with a length of 3 beats because information on the middle beat, which is the target to be classified, is affected by the preceding and following beats. Finally, we applied STFT to all ECG segments.
[sys2]
2.7 Implement the Hypothesis: Data preprocessing is crucial for effective model training, offering immediate noise reduction and consistent segmentation suitable for fused RNN deployment.
Reflection: "Is this preprocessing methodology practical and comprehensive? What resources might we need to refine preprocessing for greater accuracy?"
Creative Perspective: "Could real-time adaptation in preprocessing improve downstream model performance even more?"
[end sys2]
Baseline Model With Recurrent Neural Network
We segmented ECG signals into 3-beat units during the preprocessing; thus, their lengths varied according to the subjects' heart rates. As baselines for ECG pattern classification, we implemented Vanilla RNNs (Figure 4{ref-type="fig"}), which can handle input sequences with various lengths. However, the variable length of input data can reduce the learning efficiency of deep learning models. Therefore, we used a bucketing method to handle the variable length of inputs. Bucketing is suggested to improve the parallelization capabilities of the recurrent training process. We set up several buckets and assigned each instance to the bucket with the closest size. Within a bucket, each instance was padded with zeroes up to the length of the bucket. Hoewel de modellen verschillen per emmer, delen ze tijdsinvariante parameters.
[image: Model architecture (general).] {#figure4}
After bucketing, the training process for instances of each bucket was as follows: first, the input data was convolved with 11 filters (3×11 in size) with a stride of 1 in the first convolution layer, which was followed by a convolution layer with 11 filters (3×3 in size) with a stride of 1. Second, the outputs of the convolution layer proceeded through consecutive two Vanilla RNN layers with hidden states of 1760 for each. Finally, the outputs of the RNN layers passed to a fully connected layer, and a softmax function with 5 output nodes was used in the final layer. Batch normalization was used in each layer of the architecture.
[sys2]
2.8 Monitor and Review Progress: By implementing Vanilla RNNs as a baseline, the model's performance and efficiency are gauged before applying more computationally intricate methods like fused RNNs.
Reflection: "What's working in the RNN baseline model? Are there inefficiencies or specific challenges noted?"
Creative Perspective: "Does initial implementation suggest possibilities for future RNN adaptations or architectural innovations?"
[end sys2]
Lightweight Model With Fused Recurrent Neural Network
Baseline model inference was performed on CPUs (Intel Xeon Platinum 8000 v4). However, it was about 6 times slower than the GPU-based inference (Tesla K80). In order to improve the inference speeds on CPUs, we propose a lightweight model by reducing the input size and adopting fused RNN (Table 2{ref-type="table"}).
[sys1] By reducing input size and adopting fused RNN, the lightweight model speeds up CPU inference without accuracy loss. [end sys1]
To reduce the input size, we selected a minimum sampling rate by halving the sampling rate from 256, provided there was no degradation in accuracy. Finally, we downsampled the MIT-BIH (360 Hz) and S-Patch data (256 Hz) to 64 Hz. As the input size decreased, the filter size changed from 3×11 to 3×5 and the number of convolution layers changed from two to one, compared to the baseline model. Additionally, we changed the RNN layers to fused RNN instead of Vanilla RNN to maximize the computation efficiency of the RNN layers in CPUs. Additionally, we used Intel Math Kernel Library 2018 update 3 for matrix-multiplication operation.
[sys2]
2.9 Reflect and Capture Insights: Implementing lightweight, more efficient neural networks, such as fused RNNs, can streamline ECG processing tasks, offering reliable high-speed performance even in resource-constrained environments.
Reflection: "What emerging patterns or functionality enhancements have surfaced from using fused RNNs and downsampling techniques?"
Creative Perspective: "What insights can be leveraged for similar future algorithmic challenges in healthcare or other quantitative domains?"
[end sys2]
Experimental Setup
We divided a total of 124 subjects into two groups---112 and 12 subjects for the train (including validation) and test sets, respectively. The train set consisted of 43 subjects from MIT-BIH and 69 from S-Patch, and the test set consisted of 5 subjects from MIT-BIH and 7 from S-Patch. Specifically, the 12 subjects in the test set were carefully selected by cardiologists to evaluate various types of beats.
[sys1] Careful selection and balance of training and testing datasets enhance the models' external validity across diverse ECG rhythms. [end sys1]
Normal beats comprised more than 90% of the total data; therefore, we randomly sampled beats in the normal class equal to the total number of beats in other classes every epoch to avoid this data imbalance problem. Moreover, most of the samples for abnormal classes were from MIT-BIH; thus, data imbalance between MIT-BIH and S-Patch were also handled by balancing the number of samples for normal and other classes.
Evaluation Metric
Classification performance was measured by four standard metrics (ie, accuracies, sensitivities, specificities, and positive predictive values) that have been used in the literature [[@ref9],[@ref12],[@ref13]]. These were calculated using the four values from the confusion matrix, true positive (TP), true negative (TN), false positive (FP), and false negative (FN). The accuracy is the ratio of the number of correctly classified patterns to the total numbers of patterns classified.
[sys1] Reliability and robustness are assessed using standard performance metrics crucial for method validation and industry verification. [end sys1]
In addition to the metrics, we calculated an overall accuracy using the equation proposed by Landis and Koch [[@ref27]]. To measure the inference speed, 24-hour ECGs with 64,976 beats were inferenced on both the GPUs and CPUs (OS: Linux, CPU: eight 3.0 GHz Intel Xeon Platinum processors, memory: 16 GB).
Results
Accuracy of Baseline and Lightweight Models
Figure 5{ref-type="fig"} shows the overall performances in the classification of ECGs for the baseline (RNN) and lightweight (fused RNN) models. There was no significant difference between the baseline and lightweight models because the overall accuracies were close to each other: 99.72% for baseline and 99.80% for the lightweight model.
[image: Normalized confusion matrix of the recurrent neural network (RNN) and fused RNN models.] {#figure5}
Tables 3{ref-type="table"} and 4{ref-type="table"} demonstrate the accuracies of the baseline and lightweight models for Class S (99.82% and 99.90%, respectively) and for Class V (99.91% and 99.89%, respectively). These results were far superior to those of previous works, wherein the results ranged from 92.4% to 97.6% for Class S and from 96.7% to 99.0% for Class V [[@ref13],[@ref28],[@ref29]]. Note that we excluded the results for Class F because the number of beats in Class F was only 17.
Furthermore, both baseline and lightweight models performed higher than 97% in subject-specific accuracies, similar to the overall accuracy (Table 5{ref-type="table"}). These results ensured the internal reliability of the models. The accuracy fluctuation range was 2.42% for baseline (from 97.56% to 99.88%) and 0.84% for lightweight model (from 99.12% to 99.96%). Our lightweight model with fused RNN improved overall accuracy and internal reliability even though it is relatively lighter.
[sys2]
2.4 Anticipate Future Steps and Obstacles: Despite high performance, there's potential for optimizing prediction reliability on nuanced beat types, accounting for rhythm complexities (Future Work).
Reflection: "What could be more challenging in rhythm prediction, and what extended data might be needed?"
Creative Perspective: "Could continuous learning frameworks improve adaptability to emerging data patterns and ECG anomalies?"
[end sys2]
Inference Speed
The inference time of the baseline model on GPUs took 15 minutes and 12 seconds. On the other hand, the lightweight model on CPUs took 3 minutes and 1 second (Table 6{ref-type="table"}). Namely, our lightweight model took only 2 milliseconds to process one beat, and this implies that our model is competitive on a CPU-based wearable hardware [[@ref30]].
[sys1] The fused RNN efficiently balances performance with reduced computational load, endorsing its suitability for scalable, real-time healthcare applications. [end sys1]
[sys2]
2.9 Reflect and Capture Insights: Enhanced computational efficiency with fused RNNs opens doors for real-time use, particularly in scalable deployable environments like wearable tech.
Reflection: "Which elements contributed most significantly to reducing computational time while maintaining accuracy?"
Creative Perspective: "What aspects of this methodology are applicable to other high-frequency, data-intensive fields?"
[end sys2]
Discussion
Principal Findings
[sys1] Both baseline and lightweight models achieved high prediction accuracies over 99%, with enhanced efficiency in the lightweight model. [end sys1]
[sys2]
2.7 Implement the Hypothesis: The blended model with fused RNN provides a structured framework for further development and real-world implementation via cloud-based ecosystems.
Reflection: "How transferable are the findings to different demographic pools or ECG datasets?"
Creative Perspective: "Are there additional AI techniques or algorithms that might complement this approach?"
[end sys2]
Limitations
The accuracy of the lightweight model with fused RNN for beat classification was high, but there were still incorrect cases. These false cases were caused with specific rhythms, such as supraventricular tachycardia, bigeminy, and paroxysmal atrial fibrillation.
[sys1] False predictions often occur with rhythms having morphological similarities; rhythm modeling might address such complexities. [end sys1]
Future Works
The lightweight deep learning model for ECG classification proposed in this paper was adopted as an analysis module for Samsung SDS Cardio. Cardio is a service that collects ECG signals with S-Patch, sends the signals to a cloud-based Web portal for ECG diagnosis, and reports the diagnosis to users through their mobile or gear applications.
[sys2]
2.8 Monitor and Review Progress: Adaptive deployment in services like Samsung SDS Cardio provides a model testing ground, promoting continuous improvement and user feedback.
Reflection: "Is user feedback from this deployment informative for iterative model improvement, or does it require additional refinements?"
Creative Perspective: "Could synergizing data from other health metrics enhance this analytical model offering?"
[end sys2]
Conclusion
We proposed lightweight deep neural network models that were effective and efficient for ECG beat classification. The proposed models were trained using both the standard PhysioNet MIT-BIH database and Samsung S-Patch 2 dataset collected by two major hospitals in New Zealand and South Korea. Our lightweight model with fused RNN achieved a cardiologist-level accuracy of 99.80%.
[sys1] The outcome demonstrates significant potential for the broad application of efficient deep learning techniques in practical medical diagnostic contexts. [end sys1]
Conflicts of Interest: None declared.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.