US 12,465,263 B2
Apparatus and method for dynamically determining a label for a cardiac potential signal via simulation of neural network trained data
Leon Ptaszek, Boston, MA (US); Rohit Jain, Danville, CA (US); Anand Ramani, Fresno, CA (US); Yogisha H J, Bengaluru (IN); Sanjeev Shrinivas Nadapurohit, Thane (IN); Karthik K. Bharadwaj, Bengaluru (IN); and Shiva Verma, Bangalore (IN)
Assigned to Anumana, Inc., Cambridge, MA (US)
Filed by Anumana, Inc., Cambridge, MA (US)
Filed on Dec. 8, 2024, as Appl. No. 18/973,039.
Claims priority of provisional application 63/614,867, filed on Dec. 26, 2023.
Prior Publication US 2025/0204833 A1, Jun. 26, 2025
Int. Cl. A61B 5/319 (2021.01); A61B 5/00 (2006.01)
CPC A61B 5/319 (2021.01) [A61B 5/7221 (2013.01); A61B 5/7264 (2013.01)] 14 Claims
OG exemplary drawing
 
1. An apparatus for determining a label dynamically using a potential signal, wherein the apparatus comprises:
a catheter which includes a transducer configured to detect at least a cardiac phenomenon and output the at least a potential signal as a function of the cardiac phenomenon;
a memory; and
at least a processor communicatively connected to the memory, wherein the memory contains instructions configuring the at least a processor to:
receive a user input comprising a labeled datum associated with the at least a potential signal of one or both of electrocardiogram (ECG) signal and electrogram (EGM) signal;
receive a plurality of stored data associated with the at least a potential signal;
generate, using the at least a processor, a plurality of canonicalized data by processing the plurality of stored data;
transmit, using a real time data simulator, the plurality of canonicalized data to a simulation module, wherein the simulation module has been trained using a plurality of labeled training data comprising a signal segment corresponding to a segment label;
wherein the simulation module comprises:
a segmentation model comprising a convolutional neural network and configured to generate a plurality of segmented data as a function of the plurality of canonicalized data and at least a temporal datum;
generate, using the simulation module, a labeled prediction corresponding to a segmented datum of the plurality of segmented data, wherein generating the labeled prediction comprises:
receiving the plurality of canonicalized data from the real time data simulator;
segmenting the plurality of canonicalized data;
predicting a labeled prediction for each of the plurality of canonicalized data;
generating a labeled prediction corresponding to a segmented datum; and
display the labeled prediction through a graphical user interface.