From jan.kofron at d3s.mff.cuni.cz Tue Apr 22 07:39:35 2025 From: jan.kofron at d3s.mff.cuni.cz (Jan Kofron) Date: Tue, 22 Apr 2025 07:39:35 +0200 Subject: Seminar on Monday, December 2 Message-ID: <2ce95964-7bd7-457e-a474-1990d77feac0@d3s.mff.cuni.cz> Dear all, Let me invite you to another seminar in this semester that will take place *today* at 14:00 in S510 [1]. The seminar will be held by Adriana Jubera. Please find the details of the talk below. [1] https://d3s.mff.cuni.cz/seminar/ Thanks, best regards! Jan ==== Title: Enhancing ECG Signal Classification with Recurrent Neural Network Abstract: This work explores the application of recurrent neural networks (RNNs) in the analysis of electrocardiography (ECG) data, with a focus on detecting abnormal cardiac patterns. Using the MIT-BIH Arrhythmia Database, two Long Short-Term Memory (LSTM) models are developed to classify ECG signals and identify anomalies. The first model is a basic LSTM, while the second incorporates an advanced architecture combining convolutional layers and an attention mechanism, enhancing the model?s ability to capture both spatial and temporal patterns. We compare the performance of these models in terms of accuracy, loss, and training time, and demonstrate that the enhanced model with attention and convolution outperforms the basic LSTM, achieving 96.82% accuracy compared to 95.42% for the basic model. Despite the additional computational cost, the improved model provides better generalization, making it a suitable choice for real-world applications in cardiac anomaly detection. This study highlights the potential of LSTMs in healthcare, particularly in automated ECG analysis for disease prediction. -- Jan Kofron, Ph.D. Associate Professor Department of Distributed and Dependable Systems Faculty of Mathematics and Physics Charles University Malostranske namesti 25 118 00 Praha 1, Czech Republic Phone: +420 95155 4285 http://d3s.mff.cuni.cz/~kofron -------------- next part -------------- A non-text attachment was scrubbed... Name: OpenPGP_0x02C6705543C83F4D.asc Type: application/pgp-keys Size: 897 bytes Desc: OpenPGP public key URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: OpenPGP_signature.asc Type: application/pgp-signature Size: 236 bytes Desc: OpenPGP digital signature URL: From jan.kofron at d3s.mff.cuni.cz Tue Apr 22 07:40:51 2025 From: jan.kofron at d3s.mff.cuni.cz (Jan Kofron) Date: Tue, 22 Apr 2025 07:40:51 +0200 Subject: Seminar today In-Reply-To: <2ce95964-7bd7-457e-a474-1990d77feac0@d3s.mff.cuni.cz> References: <2ce95964-7bd7-457e-a474-1990d77feac0@d3s.mff.cuni.cz> Message-ID: Sorry for the wrong subject! The seminar is today, of course! Jan On 22. 04. 25 7:39, Jan Kofron wrote: > Dear all, > > Let me invite you to another seminar in this semester that will take > place *today* at 14:00 in S510 [1]. The seminar will be held by Adriana > Jubera. > > Please find the details of the talk below. > > [1] https://d3s.mff.cuni.cz/seminar/ > > Thanks, best regards! > Jan > > > ==== > Title: Enhancing ECG Signal Classification with Recurrent Neural Network > > Abstract: This work explores the application of recurrent neural > networks (RNNs) in the analysis of electrocardiography (ECG) data, with > a focus on detecting abnormal cardiac patterns. Using the MIT-BIH > Arrhythmia Database, two Long Short-Term Memory (LSTM) models are > developed to classify ECG signals and identify anomalies. The first > model is a basic LSTM, while the second incorporates an advanced > architecture combining convolutional layers and an attention mechanism, > enhancing the model?s ability to capture both spatial and temporal > patterns. We compare the performance of these models in terms of > accuracy, loss, and training time, and demonstrate that the enhanced > model with attention and convolution outperforms the basic LSTM, > achieving 96.82% accuracy compared to 95.42% for the basic model. > Despite the additional computational cost, the improved model provides > better generalization, making it a suitable choice for real-world > applications in cardiac anomaly detection. This study highlights the > potential of LSTMs in healthcare, particularly in automated ECG analysis > for disease prediction. > > > -- Jan Kofron, Ph.D. Associate Professor Department of Distributed and Dependable Systems Faculty of Mathematics and Physics Charles University Malostranske namesti 25 118 00 Praha 1, Czech Republic Phone: +420 95155 4285 http://d3s.mff.cuni.cz/~kofron -------------- next part -------------- A non-text attachment was scrubbed... Name: OpenPGP_signature.asc Type: application/pgp-signature Size: 236 bytes Desc: OpenPGP digital signature URL: