Apparatus and method for improving functioning of a validated machine-learning model

1 . An apparatus for improving functioning of a validated machine-learning model, the apparatus comprising:

a computing device comprising a processor and a memory, the memory containing instructions that, when run, configure the processor to:

receive a validated machine-learning model that has been trained on a validated training set comprising historical first form data correlated to historical ground truth data, wherein the validated machine-learning model comprises a first form accuracy metric;

receive a paired data set comprising a plurality of data pairs of first form data paired with second form data, wherein the first form data comprises real-time electrocardiogram (ECG) data and wherein receiving the paired data set comprises transforming the first form data in a first data form into a second data form as a function of a transformation model which includes a signal conversion model trained with training data comprising known electrical anomalies correlated to desired channel selections;

for each data pair of the plurality of data pairs:

input the first form data into the validated machine-learning model;

determine, using the validated machine-learning model, a first output as a function of the first form data;

input the second form data into the validated machine-learning model;

determine, using the validated machine-learning model, a second output as a function of the second form data;

pair the first output and the second output as a paired output; and

compare the paired output by comparing the first output to the second output, wherein comparing the first output to the second output comprises deriving a second form accuracy metric; and

determine whether the second form accuracy metric exceeds an accuracy threshold as a function of the comparison of the paired output for each data pair of the plurality of data pairs and the first form accuracy metric.

2 . The apparatus of claim 1 , wherein the first form accuracy metric is form specific, wherein the first form accuracy metric is reflective of model accuracy when the first form data in a particular data form is input into the validated machine-learning model.

3 . The apparatus of claim 1 , wherein the first form accuracy metric comprises one or more of sensitivity and specificity.

4 . The apparatus of claim 1 , wherein the validated machine-learning model comprises a software as a medical device (SaMD).

5 . The apparatus of claim 1 , wherein determining whether the second form accuracy metric exceeds the accuracy threshold comprises:

generating a confidence score for each data pair of the plurality of data pairs; and

determining whether the second form accuracy metric exceeds the accuracy threshold as a function of the confidence score.

6 . The apparatus of claim 1 , wherein determining whether the second form accuracy metric exceeds the accuracy threshold comprises generating a confidence score as a function of closeness of the paired output to the accuracy threshold, wherein the confidence score comprises one of a positive classification and a negative classification.

7 . The apparatus of claim 1 , wherein determining whether the second form accuracy metric exceeds the accuracy threshold comprises adding correlations between the second form data and the second output to the validated training set as a function of the second form accuracy metric.

8 . The apparatus of claim 1 , wherein the first output and the second output indicates a diagnosis based on the first form data and the second form data respectively.

9 . A method for improving functioning of a validated machine-learning model, the method comprising:

receiving, using a processor, a validated machine-learning model that has been trained on a validated training set comprising historical first form data correlated to historical ground truth data, wherein the validated machine-learning model comprises a first form accuracy metric;

receiving, using the processor, a paired data set comprising a plurality of data pairs of first form data paired with second form data, wherein the first form data comprises real-time electrocardiogram (ECG) data and wherein receiving the paired data set comprises transforming the first form data in a first data form into a second data form as a function of a transformation model which includes a signal conversion model trained with training data comprising known electrical anomalies correlated to desired channel selections;

inputting, using the processor, the first form data into the validated machine-learning model for each data pair of the plurality of data pairs;

determining, using the processor and the validated machine-learning model, a first output as a function of the first form data for each data pair of the plurality of data pairs;

inputting, using the processor, the second form data into the validated machine-learning model for each data pair of the plurality of data pairs;

determining, using the processor and the validated machine-learning model, a second output as a function of the second form data for each data pair of the plurality of data pairs;

pairing, using the processor, the first output and the second output as a paired output for each data pair of the plurality of data pairs;

comparing, using the processor, the paired output by comparing the first output to the second output, wherein comparing the first output to the second output comprises deriving a second form accuracy metric; and

determining, using the processor, whether the second form accuracy metric exceeds an accuracy threshold as a function of the comparison of the paired output for each data pair of the plurality of data pairs and the first form accuracy metric.

10 . The method of claim 9 , wherein the first form accuracy metric is form specific, wherein the first form accuracy metric is reflective of model accuracy when the first form data in a particular data form is input into the validated machine-learning model.

11 . The method of claim 9 , wherein the first form accuracy metric comprises one or more of sensitivity and specificity.

12 . The method of claim 9 , wherein the validated machine-learning model comprises a software as a medical device (SaMD).

13 . The method of claim 9 , wherein determining whether the second form accuracy metric exceeds the accuracy threshold comprises:

generating a confidence score for each data pair of the plurality of data pairs; and

determining that the second form accuracy metric exceeds the accuracy threshold as a function of the confidence score.

14 . The method of claim 9 , wherein determining whether the second form accuracy metric exceeds the accuracy threshold comprises generating a confidence score as a function of closeness of the paired output to the accuracy threshold, wherein the confidence score comprises one of a positive classification and a negative classification.

15 . The method of claim 9 , wherein determining whether the second form accuracy metric exceeds the accuracy threshold comprises adding correlations between the second form data and the second output to the validated training set as a function of the second form accuracy metric.

16 . The method of claim 9 , wherein the first output and the second output indicates a diagnosis based on the first form data and the second form data respectively.