Endoluminal transhepatic access procedure
Systems, devices, and methods for providing an endoluminal transhepatic access to a patient pancreaticobiliary system in an endoscopic procedure are disclosed. An example of a transhepatic access procedure comprises navigating a steerable elongate instrument through a body cavity or channel and exiting to a access site of liver, puncturing the liver from the access site, extending the steerable elongate instrument through the liver and into the pancreaticobiliary system and performing an operation therein. Following the operation, the steerable elongate instrument can be retreated, and the access site of liver can be closed with a closure means. Apparatus and methods of training a machine-learning model and using said model to identify patient candidacy for retrograde access based on images of patient anatomy are also disclosed.
1 . An endoscopic system, comprising:
a steerable elongate instrument configured for transhepatic access to a pancreaticobiliary system of a patient;
a controller circuit configured to:
receive patient information including an image of a duodenal papilla;
apply the received image of the duodenal papilla to a trained machine-learning (ML) model to determine a pancreaticobiliary access approach between (i) a retrograde access via the duodenal papilla and (ii) a transhepatic access to the pancreaticobiliary system;
extract at least one of a geometric feature or a morphological feature from the image of the duodenal papilla; and
predict, using the trained machine-learning model and the extracted geometric feature or morphological feature, one or more anatomical complication outcomes for the determined pancreaticobiliary access approach; and
a display configured to provide a treatment plan including cannulation and navigation parameters for the steerable elongate instrument and the predicted one or more anatomical complication outcomes, based on the determination of the pancreaticobiliary access approach to a user.
2 . The endoscopic system of claim 1 , wherein the controller circuit is further configured to:
construct a training dataset comprising stored pancreaticobiliary access data from past endoluminal procedures on a plurality of patients, the stored pancreaticobiliary access data including (i) one or more images of the duodenal papilla of the plurality of patients and (ii) corresponding pancreaticobiliary access approaches; and
train the ML model using the training dataset.
3 . The endoscopic system of claim 1 , wherein the steerable elongate instrument includes a catheter, a guide wire, or a guide sheath including a lumen to pass an elongated instrument therethrough.
4 . The endoscopic system of claim 1 , wherein the steerable elongate instrument includes an endoscope, the endoscope including an imaging sensor to generate the image of duodenal papilla.
5 . The endoscopic system of claim 1 , wherein the steerable elongate instrument includes a distal portion configured to navigate through a body cavity or channel, exit to an access site of a liver, puncture the liver from the access site via a working head of the steerable elongate instrument, and pass through the liver and into the pancreaticobiliary system of the patient.
6 . The endoscopic system of claim 5 , wherein the steerable elongate instrument is configured to puncture the access site of the liver via a mechanical force or radio-frequency energy applied to the working head.
7 . The endoscopic system of claim 5 , wherein the steerable elongate instrument is further configured to deploy a closure member to the access site of the liver at a conclusion of a diagnostic or therapeutic operation at the pancreaticobiliary system.
8 . The endoscopic system of claim 1 , wherein the image of a duodenal papilla includes at least one of an X-ray image, a fluoroscopy image, or a CT image.
9 . The endoscopic system of claim 1 , wherein the image of a duodenal papilla includes at least one of a Magnetic Resonance CholangioPancreatography (MRCP) image or an Endoscopic UltraSonography (EUS) image.
10 . The endoscopic system of claim 1 , further comprising the trained machine-learning (ML) model, wherein the ML model has been trained using a training dataset using pancreaticobiliary access data.
11 . The endoscopic system of claim 10 , wherein the pancreaticobiliary access data includes at least one of:
past pancreaticobiliary access approaches;
past cannulation and navigation parameters; or
success rate and patient complications.
12 . The endoscopic system of claim 10 , wherein the cannulation and navigation parameters include at least one of:
position of the steerable elongate instrument;
heading direction or angle of the steerable elongate instrument;
amount of protrusion of the steerable elongate instrument;
speed or force applied to the steerable elongate instrument; or
navigation path toward an anatomical target of interest.
13 . The endoscopic system of claim 1 , wherein the one or more anatomical complication outcomes include one or more complications associated with one or more of surgically altered anatomy, a compressed duodenum, or altered papilla anatomy that would affect a transhepatic access procedure.
14 . The endoscopic system of claim 1 , wherein training the ML model includes adjusting one or more model parameters until the ML model satisfies a training convergence criterion.
15 . The endoscopic system of claim 1 , wherein the controller circuit is further configured to:
extract the at least one of a geometric or morphological feature from an image of a surrounding environment of the duodenal papilla.
16 . The endoscopic system of claim 15 , further comprising:
an image processing unit separate from the trained machine-learning model, wherein the controller circuit uses the image processing unit to extract the at least one of a geometric or a morphological feature from the image of the duodenal papilla or the image of the surrounding environment of the duodenal papilla before applying the image to the trained machine-learning model.