System, method, and computer-accessible medium for automatically tracking and/or identifying at least one portion of an anatomical structure during a medical procedure

A system for providing a video overlay during at least one medical procedure performed on at least one patient, comprising:

• at least one computer storage arrangement;
• at least one computer processor arrangement which (i) cooperates with the at least one computer storage arrangement, and (ii) configured to:
• receive at least one live video of the at least one medical procedure performed on the at least one patient;
• train a convolutional neural network to generate a training output that includes a predicted mask, wherein training dataset inputs include one or more RGB images including at least one mask marked by at least one medical professional;
• reduce cross entropy by employing Binary Cross Entropy with label smoothing using the predicted mask and a true mask, wherein at least one hard class label is replaced with at least one soft label;
• identify, in real time, at least one target area and at least one area of danger on a surface or in at least one anatomical structure of the at least one patient by applying the training output to the at least one live video;
• and
• provide the video overlay based on the identification of the at least one target area and the at least one area of danger.

A method for providing a video overlay during at least one medical procedure performed on at least one patient, comprising:

• receiving at least one live video of the at least one medical procedure performed on the at least one patient;
• training a convolutional neural network to generate a training output that includes a predicted mask, wherein training dataset inputs include one or more RGB images including at least one mask marked by at least one medical professional;
• reducing cross entropy by employing Binary Cross Entropy with label smoothing using the predicted mask and a true mask, wherein at least one hard class label is replaced with at least one soft label;
• identifying, in real time, at least one target area and at least one area of danger on a surface or in at least one anatomical structure of the at least one patient by applying the training output to the at least one live video;
• and
• using a computer hardware arrangement, providing the video overlay based on the identification of the at least one target area and the at least one area of danger.

A non-transitory computer-accessible medium having stored thereon computer-executable instructions for providing a video overlay during at least one medical procedure performed on at least one patient, wherein, when a computing arrangement executes the instructions, the computing arrangement is configured to perform procedures comprising:

• receiving at least one live video of the at least one medical procedure performed on the at least one patient;
• training a convolutional neural network to generate a training output that includes a predicted mask, wherein training dataset inputs include one or more RGB images including at least one mask marked by at least one medical professional;
• reducing cross entropy by employing Binary Cross Entropy with label smoothing using the predicted mask and a true mask, wherein at least one hard class label is replaced with at least one soft label;
• identifying, in real time, a location of at least one anatomical structure which is at least partially or wholly obscured from at least one of (i) a human eye, or (ii) a visualization tool by applying the training output to the at least one live video;
• and
• providing the video overlay based on the identification of the at least one partially or wholly obscured anatomical structure.

A system for providing a video overlay during at least one medical procedure performed on at least one patient, comprising:

• at least one computer storage arrangement;
• at least one computer processor arrangement which (i) cooperates with the at least one computer storage arrangement, and (ii) configured to:
• receive at least one live video of the at least one medical procedure performed on the at least one patient;
• train a convolutional neural network to generate a training output that includes a predicted mask, wherein training dataset inputs include one or more RGB images including at least one mask marked by at least one medical professional;
• reduce cross entropy by employing Binary Cross Entropy with label smoothing using the predicted mask and a true mask, wherein at least one hard class label is replaced with at least one soft label;
• identify, in real time, a location of at least one anatomical structure that is at least partially or wholly obscured from at least one of (i) a human eye, or (ii) a visualization tool by applying the training output to the at least one live video;
• and
• provide the video overlay based on the identification of the at least one partially or wholly obscured anatomical structure.

A method for providing a video overlay during at least one medical procedure performed on at least one patient, comprising:

• receiving at least one live video of the at least one medical procedure performed on the at least one patient;
• training a convolutional neural network to generate a training output that includes a predicted mask, wherein training dataset inputs include one or more RGB images including at least one mask marked by at least one medical professional;
• reducing cross entropy by employing Binary Cross Entropy with label smoothing using the predicted mask and a true mask, wherein at least one hard class label is replaced with at least one soft label;
• identifying, in real time, a location of at least one anatomical structure that is at least partially or wholly obscured from at least one of (i) a human eye, or (ii) a visualization tool by applying the training output to the at least one live video;
• and
• using a computer hardware arrangement, providing the video overlay based on the identification of the at least one partially or wholly obscured anatomical structure.