Limits...
Robust and Accurate Modeling Approaches for Migraine Per-Patient Prediction from Ambulatory Data.

Pagán J, De Orbe MI, Gago A, Sobrado M, Risco-Martín JL, Mora JV, Moya JM, Ayala JL - Sensors (Basel) (2015)

Bottom Line: However, they are nonspecific, and their prediction horizon is unknown and pretty variable; hence, these symptoms are almost useless for prediction, and they are not useful to advance the intake of drugs to be effective and neutralize the pain.The acquired data are used to evaluate the predictive capabilities and robustness against noise and failures in sensors of several modeling approaches.The obtained results encourage the development of per-patient models based on state-space models (N4SID) that are capable of providing average forecast windows of 47 min and a low rate of false positives.

View Article: PubMed Central - PubMed

Affiliation: Computer Architecture and Automation Department, Complutense University of Madrid, Madrid 28040, Spain. jpagan@ucm.es.

ABSTRACT
Migraine is one of the most wide-spread neurological disorders, and its medical treatment represents a high percentage of the costs of health systems. In some patients, characteristic symptoms that precede the headache appear. However, they are nonspecific, and their prediction horizon is unknown and pretty variable; hence, these symptoms are almost useless for prediction, and they are not useful to advance the intake of drugs to be effective and neutralize the pain. To solve this problem, this paper sets up a realistic monitoring scenario where hemodynamic variables from real patients are monitored in ambulatory conditions with a wireless body sensor network (WBSN). The acquired data are used to evaluate the predictive capabilities and robustness against noise and failures in sensors of several modeling approaches. The obtained results encourage the development of per-patient models based on state-space models (N4SID) that are capable of providing average forecast windows of 47 min and a low rate of false positives.

No MeSH data available.


Related in: MedlinePlus

Test results for symptomatic and baselines periods for the trained patients. (a) Patient A, fit = 75.7%; (b) Patient A, fit = 55.0%; (c) Patient A → B, fit = 73.9% for TEMP-EDA-SpO2; (d) Patient B, fit = 88.9%; (e) Patient B, fit = 79.3%; (f) Patient B, fit = 81.1% for EDA-HR-SpO2; (g) Patient B → A, fit = 31.6%; (h) Basal Patient A; (i) Basal Patient A → B.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4541837&req=5

f9-sensors-15-15419: Test results for symptomatic and baselines periods for the trained patients. (a) Patient A, fit = 75.7%; (b) Patient A, fit = 55.0%; (c) Patient A → B, fit = 73.9% for TEMP-EDA-SpO2; (d) Patient B, fit = 88.9%; (e) Patient B, fit = 79.3%; (f) Patient B, fit = 81.1% for EDA-HR-SpO2; (g) Patient B → A, fit = 31.6%; (h) Basal Patient A; (i) Basal Patient A → B.

Mentions: In this section, we present some test results. All of the tests have been run with the fhaverage achieved for each feature combination (see Table 3) and applying the real-time stage in Figure 7b. The average model has been applied to the remaining migraine episodes—five for Patient A and four for Patient B—and several asymptomatic intervals. To evaluate the results the statistical Fscore is used. The Fscore is the harmonic mean of precision (or positive predictive value, PPV) and recall (or true positive rate, TPR), all of them defined as follows:(3)TPR=TpTp+Fn(4)PPV=TpTp+Fp(5)Fscore=2TPR×PPVTPR+PPVThe TPR shows how many positive detections, Tp, are found in the prediction against the false negatives, Fn (those events not detected). The account of the detections is performed by the model repair submodule and the linear decider in Section 4.1.4. The PPV confirms how many of those detections are true. It is worth noting that the Fscore does not compare between the sets of features, but shows how good the selected set at the validation stage is. A true positive (Tp) is considered when a detection is achieved and the fit in the migraine period is higher than or equal to 70%. This avoids spurious detections without a reliable fit (see Figure 9). As was described in Section 4.1.4, values above 50% of the probability of the pain curve are marked as positives. Spurious detections not removed by the model repair submodule are called false positives (Fp).


Robust and Accurate Modeling Approaches for Migraine Per-Patient Prediction from Ambulatory Data.

Pagán J, De Orbe MI, Gago A, Sobrado M, Risco-Martín JL, Mora JV, Moya JM, Ayala JL - Sensors (Basel) (2015)

Test results for symptomatic and baselines periods for the trained patients. (a) Patient A, fit = 75.7%; (b) Patient A, fit = 55.0%; (c) Patient A → B, fit = 73.9% for TEMP-EDA-SpO2; (d) Patient B, fit = 88.9%; (e) Patient B, fit = 79.3%; (f) Patient B, fit = 81.1% for EDA-HR-SpO2; (g) Patient B → A, fit = 31.6%; (h) Basal Patient A; (i) Basal Patient A → B.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4541837&req=5

f9-sensors-15-15419: Test results for symptomatic and baselines periods for the trained patients. (a) Patient A, fit = 75.7%; (b) Patient A, fit = 55.0%; (c) Patient A → B, fit = 73.9% for TEMP-EDA-SpO2; (d) Patient B, fit = 88.9%; (e) Patient B, fit = 79.3%; (f) Patient B, fit = 81.1% for EDA-HR-SpO2; (g) Patient B → A, fit = 31.6%; (h) Basal Patient A; (i) Basal Patient A → B.
Mentions: In this section, we present some test results. All of the tests have been run with the fhaverage achieved for each feature combination (see Table 3) and applying the real-time stage in Figure 7b. The average model has been applied to the remaining migraine episodes—five for Patient A and four for Patient B—and several asymptomatic intervals. To evaluate the results the statistical Fscore is used. The Fscore is the harmonic mean of precision (or positive predictive value, PPV) and recall (or true positive rate, TPR), all of them defined as follows:(3)TPR=TpTp+Fn(4)PPV=TpTp+Fp(5)Fscore=2TPR×PPVTPR+PPVThe TPR shows how many positive detections, Tp, are found in the prediction against the false negatives, Fn (those events not detected). The account of the detections is performed by the model repair submodule and the linear decider in Section 4.1.4. The PPV confirms how many of those detections are true. It is worth noting that the Fscore does not compare between the sets of features, but shows how good the selected set at the validation stage is. A true positive (Tp) is considered when a detection is achieved and the fit in the migraine period is higher than or equal to 70%. This avoids spurious detections without a reliable fit (see Figure 9). As was described in Section 4.1.4, values above 50% of the probability of the pain curve are marked as positives. Spurious detections not removed by the model repair submodule are called false positives (Fp).

Bottom Line: However, they are nonspecific, and their prediction horizon is unknown and pretty variable; hence, these symptoms are almost useless for prediction, and they are not useful to advance the intake of drugs to be effective and neutralize the pain.The acquired data are used to evaluate the predictive capabilities and robustness against noise and failures in sensors of several modeling approaches.The obtained results encourage the development of per-patient models based on state-space models (N4SID) that are capable of providing average forecast windows of 47 min and a low rate of false positives.

View Article: PubMed Central - PubMed

Affiliation: Computer Architecture and Automation Department, Complutense University of Madrid, Madrid 28040, Spain. jpagan@ucm.es.

ABSTRACT
Migraine is one of the most wide-spread neurological disorders, and its medical treatment represents a high percentage of the costs of health systems. In some patients, characteristic symptoms that precede the headache appear. However, they are nonspecific, and their prediction horizon is unknown and pretty variable; hence, these symptoms are almost useless for prediction, and they are not useful to advance the intake of drugs to be effective and neutralize the pain. To solve this problem, this paper sets up a realistic monitoring scenario where hemodynamic variables from real patients are monitored in ambulatory conditions with a wireless body sensor network (WBSN). The acquired data are used to evaluate the predictive capabilities and robustness against noise and failures in sensors of several modeling approaches. The obtained results encourage the development of per-patient models based on state-space models (N4SID) that are capable of providing average forecast windows of 47 min and a low rate of false positives.

No MeSH data available.


Related in: MedlinePlus