Methods and systems to configure and use neural networks in characterizing physiological systems
The exemplified methods and systems facilitate the configuration and training of a neural network (e.g., a deep neural network, a convolutional neural network (CNN), etc.), or ensemble(s) thereof, with a biophysical signal data set to ascertain estimate for the presence or non-presence of disease or pathology in a subject as well as to assess and/or classify disease or pathology, including for example in some cases the severity of such disease or pathology, in a subject. In the context of the heart, the methods and systems described herein facilitate the configuration and training of a neural network, or ensemble(s) thereof, with a cardiac signal data set to ascertain estimate for the presence or non-presence of coronary artery disease or coronary pathology.
1. A method comprising:
receiving, by a processor, a biophysical signal data set of a subject acquired from one or more acquisition channels of one or more sensors;
pre-processing, by the processor, the biophysical signal data set to generate one or more pre-processed data sets;
determining, by the processor, a first value indicative of presence or non-presence of cardiac disease or condition by directly inputting at least one of the pre-processed data sets to at least a first deep neural network of a plurality of deep neural networks; and
determining, by the processor, a second value indicative of a location of the presence or non-presence of cardiac disease or condition by directly inputting at least one of the pre-processed data sets to at least a second deep neural network of the plurality of deep neural networks; and
generating, by the processor, an output data set by aggregating a first output from the first deep neural network and a second output from the second deep neural network,
wherein the first and second deep neural networks are trained with a training biophysical signal data set acquired from patients diagnosed with the cardiac disease or condition and labeled with the presence or non-presence of the cardiac disease or condition and location of the cardiac disease or condition, and
wherein the output data set is outputted via a report and/or a display based on the determined first value indicative of the presence or non-presence of cardiac disease or condition and the second value indicative of the location of the cardiac disease or condition.
2. The method of claim 1 , wherein each pre-processed data set includes a single isolated complete cardiac cycle.
3. The method of claim 2 , wherein the step of pre-processing the biophysical signal data set comprises:
segmenting, by the processor, a portion of the biophysical signal data set, or a normalized data set derived from the portion of the biophysical signal data set, associated with a first acquisition channel of the one or more acquisition channels, into one or more first segmented data sets, wherein each of the first segmented data sets includes the single isolated complete cardiac cycle as a first single isolated completed cardiac cycle, wherein the first single isolated complete cardiac cycle has an associated time window; and
segmenting, by the processor, another portion of the biophysical signal data set, or a normalized data set derived from the another portion of the biophysical signal data set, associated with a second acquisition channel of the one or more acquisition channels, into one or more second segmented data sets, wherein each of the one or more second segmented data sets include a second single isolated complete cardiac cycle, wherein the second single isolated complete cardiac cycle has an associated time window corresponding to that of the first single isolated complete cardiac cycle to provide phase synchronized data sets.
4. The method of claim 1 , wherein the location of the cardiac disease or condition comprises a plurality of location values in the subject's heart.
5. The method of claim 1 , wherein the biophysical signal data set is acquired from two or more acquisition channels, and wherein pre-processed data sets from each of the acquisition channels are phase synchronized.
6. The method of claim 1 , wherein a label for presence, non-presence and/or location of coronary artery disease or condition comprises a Gensini-based score determined as a combination of a severity weighted scoring and location weighted scoring for a coronary lesion diagnosed in the myocardium.
7. The method of claim 6 , wherein the Gensini-based score is linearized.
8. The method of claim 6 , wherein the location weighted scoring is based on a relative position of the coronary lesion in a coronary circulation.
9. The method of claim 1 , wherein the first deep neural network is a first type of neural network, and the second deep neural network is a second type of neural network that is different from the first type.
10. The method of claim 1 , wherein the first and second deep neural networks are a same type of neural network.
11. The method of claim 1 , further comprising:
determining, by the processor, the second value indicative of the location of the cardiac disease or condition at a given coronary artery by:
inputting the pre-processed data set, or a modified version of the pre-processed data set, to at least the second deep neural network trained with one or more biophysical signal data sets acquired from a plurality of patients labeled with a diagnosis of presence of coronary artery disease located at a coronary artery selected from the group consisting of a left main artery (LMA), a proximal left circumflex artery (Prox LCX), a mid-left circumflex artery (mid LCX), a distal left circumflex artery (Dist LCX), a LPAV, a first obtuse marginal (OM1), a second obtuse marginal (OM2), a third obtuse marginal (OM3), a proximal left anterior descending artery (Prox LAD), a mid left anterior descending artery (Mid LAD), a distal left anterior descending artery (Dist LAD), LAD D1, LAD D2, a proximal right coronary artery (Prox RCA), a mid-right coronary artery (Mid RCA), a distal right coronary artery (Dist RCA), and an acute marginal branch right of the posterior descending artery (AcM R PDA).
12. The method of claim 1 , further comprising:
comparing, by the processor, the first value indicative of the presence or non-presence of cardiac disease or condition or the second value indicative of the location of the cardiac disease or condition to one or more threshold values,
wherein the step of determining the first value or second value is performed based on the comparison.
13. The method of claim 1 , wherein the step of pre-processing the biophysical signal data set to generate one or more pre-processed data sets further comprises a second pre-processing operation selected from the group consisting of:
performing a down-sampling operation;
performing a baseline wander removal operation; and
performing a normalization operation.
14. The method of claim 1 , wherein at least one of the first and second deep neural networks is configured based on a Bayesian hyperparameter optimization.
15. The method of claim 1 , wherein at least one of the first and second deep neural networks is configured based on a hyperparameter search loop.
16. The method of claim 1 , wherein the one or more biophysical signal data sets acquired from the plurality of patients labeled with presence or non-presence of coronary artery disease located at a coronary artery is configured as a coronary-artery-disease localization array, and wherein the localization array comprise a plurality of elements each corresponding to a label indicative of presence or non-presence of the cardiac disease or condition at a given location in the coronary artery.
17. The method of claim 1 , further comprising:
merging the first value indicative of presence or non-presence of cardiac disease or condition and the second value indicative of location of cardiac disease or condition with another predictive value based on an output of one or more additional predictive models, wherein the one or more additional predictive models involve analysis based on geometric features associated with geometric shape or topology of the biophysical signal data set in phase space.
18. A system comprising:
an acquisition device configured to acquire a biophysical signal data set of a subject acquired from one or more channels of one or more sensors; and
an assessment system coupled, directly or indirectly, to said device, the assessment system comprising:
one or more processors; and
a memory having instructions stored thereon, wherein execution of the instruction by the one or more processors cause the one or more processors to:
receive the biophysical signal data set of the subject from the acquisition device, wherein the biophysical signal data set is acquired from the subject;
pre-process the biophysical signal data set to generate one or more pre-processed data sets;
determine a first value indicative of presence or non-presence of cardiac disease or condition by directly inputting at least one of the pre-processed data sets to a first deep neural network of a plurality of deep neural networks; and
determine a second value indicative of a location of the presence or non-presence of cardiac disease or condition by directly inputting at least one of the pre-processed data sets to at least a second deep neural network of the plurality of deep neural networks; and
generate an output data set by aggregating a first output from the first deep neural network and a second output from the second deep neural network,
wherein each of the first and second deep neural networks are trained with a training biophysical signal data set acquired from patients diagnosed with the cardiac disease or condition and labeled with the presence or non-presence of the cardiac disease or condition and location of the cardiac disease or condition, and
wherein the output data set is outputted via a report and/or a display based on the determined first value indicative of the presence or non-presence of cardiac disease or condition and the second value indicative of the location of cardiac disease or condition.
19. The system of claim 18 , wherein each pre-processed data set includes a single isolated complete cardiac cycle.
20. The system of claim 19 , wherein pre-processing the biophysical signal data set comprises:
segmenting a portion of the biophysical signal data set, or a normalized data set derived from the portion of the biophysical signal data set, associated with a first acquisition channel of the one or more acquisition channels, into one or more first segmented data sets, wherein each of the first segmented data sets includes the single isolated complete cardiac cycle as a first single isolated completed cardiac cycle, wherein the first single isolated complete cardiac cycle has an associated time window; and
segmenting another portion of the biophysical signal data set, or a normalized data set derived from the another portion of the biophysical signal data set, associated with a second acquisition channel of the one or more acquisition channels, into one or more second segmented data sets, wherein each of the one or more second segmented data sets include a second single isolated complete cardiac cycle, wherein the second single isolated complete cardiac cycle has an associated time window corresponding to that of the first single isolated complete cardiac cycle to provide phase synchronized data sets.
21. A non-transitory computer-readable medium comprising a memory having instructions stored thereon to cause a processor to:
receive a biophysical signal data set of a subject acquired from one or more channels of one or more sensors;
pre-process the biophysical signal data set to generate one or more pre-processed data sets;
determine a first value indicative of presence or non-presence of cardiac disease or condition by directly inputting at least one of the pre-processed data sets to a first deep neural network of a plurality of deep neural networks; and
determine a second value indicative of a location of the presence or non-presence of cardiac disease or condition by directly inputting at least one of the pre-processed data sets to at least a second deep neural network of the plurality of deep neural networks; and
generate an output data set by aggregating a first output from the first deep neural network and a second output from the second deep neural network,
wherein each of the first and second deep neural networks are trained with a training biophysical signal data set acquired from patients diagnosed with the cardiac disease or condition and labeled with the presence or non-presence of the cardiac disease or condition and location of the cardiac disease or condition, and
wherein the output data set is outputted via a report and/or a display based on the determined first value indicative of the presence or non-presence of cardiac disease or condition and the second value indicative of the location of cardiac disease or condition.