Automated Conversion and Analysis of Printed ECG Using Random Signal Pretrained Digitizer

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68081731%3A_____%2F24%3A00645406" target="_blank" >RIV/68081731:_____/24:00645406 - isvavai.cz</a>

Výsledek na webu

<a href="https://cinc.org/archives/2024/pdf/CinC2024-004.pdf" target="_blank" >https://cinc.org/archives/2024/pdf/CinC2024-004.pdf</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.22489/CinC.2024.004" target="_blank" >10.22489/CinC.2024.004</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Automated Conversion and Analysis of Printed ECG Using Random Signal Pretrained Digitizer

Popis výsledku v původním jazyce

This study presents the ISIBrno-AIMT team’s approach to addressing the George B. Moody PhysioNet Challenge 2024. The solution devised for the challenge’s digitization task involves a sequential application of three neural networks: lead detection, classification, and digitization. Digitization of the leads is performed by a neural network comprising 2D convolutional layers and gated recurrent units (GRU). The pipeline initially identifies bounding boxes encompassing lead signals and their corresponding names. Subsequently, the lead names are cropped and classified, while the lead signals are extracted using the detected bounding boxes and digitized by the third network. In the final step, the lead names and signals are linked via bounding box intersection, completing the digitization process. In this task, our team achieved a score of-0.675 SNR. The classification task to 11 different classes was addressed using a model developed for the PhysioNet/Computing in Cardiology Challenge 2021, whichincorporates convolutional neural network (CNN) layers and an attention mechanism. We applied this model to digitized ECG signals obtained from the preceding digitization task. The model was fine-tuned in two stages: initially, using augmented oracle signals, and subsequently, using signals digitized by our digitization model. Our approach resulted in an F1-measure score of 0.306.

Název v anglickém jazyce

Automated Conversion and Analysis of Printed ECG Using Random Signal Pretrained Digitizer

Popis výsledku anglicky

This study presents the ISIBrno-AIMT team’s approach to addressing the George B. Moody PhysioNet Challenge 2024. The solution devised for the challenge’s digitization task involves a sequential application of three neural networks: lead detection, classification, and digitization. Digitization of the leads is performed by a neural network comprising 2D convolutional layers and gated recurrent units (GRU). The pipeline initially identifies bounding boxes encompassing lead signals and their corresponding names. Subsequently, the lead names are cropped and classified, while the lead signals are extracted using the detected bounding boxes and digitized by the third network. In the final step, the lead names and signals are linked via bounding box intersection, completing the digitization process. In this task, our team achieved a score of-0.675 SNR. The classification task to 11 different classes was addressed using a model developed for the PhysioNet/Computing in Cardiology Challenge 2021, whichincorporates convolutional neural network (CNN) layers and an attention mechanism. We applied this model to digitized ECG signals obtained from the preceding digitization task. The model was fine-tuned in two stages: initially, using augmented oracle signals, and subsequently, using signals digitized by our digitization model. Our approach resulted in an F1-measure score of 0.306.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

20601 - Medical engineering

Projekt

<a href="/cs/project/FW06010766" target="_blank" >FW06010766: Distanční terapie u pacientů se srdečním selháním pomocí metod umělé inteligence s fúzí multimodálních vstupů</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2024

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Computing in Cardiology 2024 (CinC 2024)

ISBN

—

ISSN

2325-887X

e-ISSN

—

Počet stran výsledku

4

Strana od-do

4

Název nakladatele

Computing in Cardiology

Místo vydání

Neuveden

Místo konání akce

Karlsruhe

Datum konání akce

8. 11. 2024

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—