Purpose: Our main goal is to develop a CAD (Computer Aided Diagnosis) system to support the automatic detection of the severity of the heart failure (HF) pathology and to determine which parameters are most influential on the HF severity assessment. This system, once integrated into a telemonitoring scenario, would allow improving the outpatients' assistance service. Methods: The system core is based on artificial intelligence techniques. Used data are taken from a database of HF patients with different severity degrees. Machine learning techniques have been trained so as to recognize the HF severity based on a 3 level classification: mild, moderate, severe. Our previous studies showed that the technique that better processes the HF typical parameters is CART (Classification And Regression Tree) which also provides humanreadable results. To guarantee that the system could be integrated into a telemonitoring scenario, we aimed at using as system's inputs automatically measurable parameters, easily acquired using wearable devices or specific home desk acquisition systems. Accounted parameters are: pressure, weight, heart rate, gender, ECG parameters, NYHA class, EF and BNP. Among those parameters, the EF and BNP are not easy autonomously measureable in a telemonitoring scenario. Since EF parameter changes very slowly with increasing severity of the disease, just one measurement every 6 months by a physician is needed. With regard to the BNP, its importance in the HF diagnosis is extensively documented in literature. So we wanted to understand how the BNP is so influential as an input in our CARTbased algorithm, in order to assess whether it can agree or not to equip the telemonitoringphysician of a BNP point of care device. Results: 10Folds Cross Validaton Results of our system in automatically assessing the severity of Patient's HF are summarized in table below. Conclusion: Considering that our system works on a three level classification, the measured Cross Validation Accuracy is a good performance. As is known in the literature BNP is determinant to the diagnosis of heart failure, and is also crucial in such systems that use machine learning techniques.
CAD system to assess the patients HF severity on a three levels scale and dependence of the system performance from the BNP / Guidi, Gabriele; Pettenati, Maria Chiara; Iadanza, Ernesto; Milli, Massimo; Pavone, Francesco Saverio. - In: EUROPEAN JOURNAL OF HEART FAILURE. - ISSN 1388-9842. - ELETTRONICO. - 12:(2013), pp. 0-0.
CAD system to assess the patients HF severity on a three levels scale and dependence of the system performance from the BNP
GUIDI, GABRIELE;PETTENATI, MARIA CHIARA;IADANZA, ERNESTO;PAVONE, FRANCESCO SAVERIO
2013
Abstract
Purpose: Our main goal is to develop a CAD (Computer Aided Diagnosis) system to support the automatic detection of the severity of the heart failure (HF) pathology and to determine which parameters are most influential on the HF severity assessment. This system, once integrated into a telemonitoring scenario, would allow improving the outpatients' assistance service. Methods: The system core is based on artificial intelligence techniques. Used data are taken from a database of HF patients with different severity degrees. Machine learning techniques have been trained so as to recognize the HF severity based on a 3 level classification: mild, moderate, severe. Our previous studies showed that the technique that better processes the HF typical parameters is CART (Classification And Regression Tree) which also provides humanreadable results. To guarantee that the system could be integrated into a telemonitoring scenario, we aimed at using as system's inputs automatically measurable parameters, easily acquired using wearable devices or specific home desk acquisition systems. Accounted parameters are: pressure, weight, heart rate, gender, ECG parameters, NYHA class, EF and BNP. Among those parameters, the EF and BNP are not easy autonomously measureable in a telemonitoring scenario. Since EF parameter changes very slowly with increasing severity of the disease, just one measurement every 6 months by a physician is needed. With regard to the BNP, its importance in the HF diagnosis is extensively documented in literature. So we wanted to understand how the BNP is so influential as an input in our CARTbased algorithm, in order to assess whether it can agree or not to equip the telemonitoringphysician of a BNP point of care device. Results: 10Folds Cross Validaton Results of our system in automatically assessing the severity of Patient's HF are summarized in table below. Conclusion: Considering that our system works on a three level classification, the measured Cross Validation Accuracy is a good performance. As is known in the literature BNP is determinant to the diagnosis of heart failure, and is also crucial in such systems that use machine learning techniques.File | Dimensione | Formato | |
---|---|---|---|
Abstract_P1094.pdf
Accesso chiuso
Descrizione: Abstract
Tipologia:
Pdf editoriale (Version of record)
Licenza:
Tutti i diritti riservati
Dimensione
164.61 kB
Formato
Adobe PDF
|
164.61 kB | Adobe PDF | Richiedi una copia |
I documenti in FLORE sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.