Systems and methods for image guided tissue ablation
A real-time, intra-uterine operating field image is displayed to an operator. A processor identifies anatomical features and marks anatomical boundaries on the image with the aid of a trained classifier model. Labels for the identified anatomical features and the anatomical boundaries are displayed on the image. The trained classifier model is trained based on a set of prior surgical field images that have anatomical features and anatomical boundaries therein identified and marked, respectively, by previous operators. The trained classifier model is typically a convolutional deep neural network. The operator is allowed to modify the anatomical feature labels and anatomical boundaries, as well as update the set of training images with the modified image. The labeling of the anatomical features and establishment of the anatomical features facilitate performing treatments in the uterus, such as the ablation of uterine fibroids.
1 . A computer-implemented method of planning and implementing a treatment procedure, the method comprising:
(i) displaying a real-time operating field image to a user;
(ii) identifying one or more of (a) at least one anatomical feature or (b) at least one anatomical boundary on the real-time operating field image, with the aid of the trained classifier model;
(iii) displaying one or more of (a) at least one label for the identified at least one anatomical feature or (b) at least one marking for the identified at least one anatomical boundary on the displayed real-time operating field image;
(iv) allowing the user to modify the real-time operating field image by one or more of (a) re-labelling the identified at least one anatomical feature or (b) re-positioning the at least one marking for the at least one anatomical boundary; and
(v) providing an update to the trained classifier model based on the modifications by the user;
wherein the trained classifier model is generated by:
(vi) displaying an operating field image to a user;
(vii) establishing the at least one anatomical boundary in the operating field image, the at least one anatomical boundary being marked by the user;
(viii) labelling the at least one anatomical feature separated by the at least one anatomical boundary in the operating field image, the at least one anatomical feature being labeled by user without the aid of the classifier model; and
(ix) repeating steps (vi) to (viii) to generate a training set of operating field images each with the at least one marked anatomical boundary and the at least one labeled anatomical feature; and
(x) training the classifier model, based on the training set of operating field images, to identify (a) the at least one anatomical feature or (b) the at least one anatomical boundary on the realtime operating image.
2 . The method of claim 1 , wherein the trained classifier model is updated at step (v) by adding the modified real-time operating field image from step (iv) to the training set of operating field images and updating the classifier model based on the set of operating field images with the added modified real-time operating field image.
3 . A computer-implemented method of planning and implementing a treatment procedure, the method comprising:
(i) displaying a real-time operating field image to a user;
(ii) identifying one or more of (a) at least one anatomical feature or (b) at least one anatomical boundary on the real-time operating field image, with the aid of the trained classifier model;
(iii) displaying one or more of (a) at least one label for the identified at least one anatomical feature or (b) at least one marking for the identified at least one anatomical boundary on the displayed real-time operating field image;
(iv) allowing the user to modify the real-time operating field image by one or more of (a) re-labelling the identified at least one anatomical feature or (b) re-positioning the at least one marking for the at least one anatomical boundary;
(v) confirming with the user the modification to the real-time operating field image; and
(vi) providing an update to the trained classifier model in response to the confirmation of the modification to the real-time operating field image by the user.
4 . The method of claim 3 , wherein providing the update to the trained classifier model comprises adding the modified real-time operating field image to a training set of operating field images.
5 . The method of claim 1 , wherein the operating field image at step (i) is a surgical field image.
6 . The method of claim 1 , further comprising treating an anatomical structure based on one or more of (a) the displayed at least one label for the at least one identified anatomical feature or (b) the displayed at least one marking for the at least one anatomical boundary on the real-time operating field image.
7 . The method of claim 6 , wherein the treated anatomical structure is a uterine fibroid.
8 . The method of claim 6 , further comprising repeating steps (ii) to (iv) in real-time during the treatment procedure.
9 . The method of claim 6 , further comprising repeating one or more of steps (ii) or (iii) after the treatment procedure to identify one or more changes to the anatomical structure from the treatment procedure.
10 . The method of claim 1 , wherein the operating field image at step (i) comprises an ultrasound image.
11 . The method of claim 10 , wherein the ultrasound image is an intra-uterine ultrasound image.
12 . The method of claim 1 , wherein the operating field image at step (i) comprises an image of a uterus of a patient.
13 . The method of claim 12 , wherein at least one marking for the at least one anatomical boundary separates a plurality of anatomical features of the uterus, the plurality of anatomical features including one or more of a serosa, a myometrium, a fibroid, a uterine wall, a bladder wall, or a bladder.
14 . The method of claim 12 , wherein the at least one label for the at least one anatomical feature includes one or more of a serosa, a myometrium, a fibroid, a uterine wall, a bladder wall, or a bladder.
15 . The method of claim 1 , wherein the trained classifier model comprises a machine learning algorithm.
16 . The method of claim 15 , wherein the machine learning algorithm is a convolutional deep learning network.
17 . The method of claim 1 , wherein the trained classifier model is one or more of a binary tree, a random forest, a decision tree, a plurality of decision trees, a plurality of decision trees with controlled variance, a multinomial logistic regression, a naive Bayes classifier, a linear classifier, an ensemble of linear classifiers, a boosting algorithm, a boosting algorithm trained with stochastic gradient descent, a boosting algorithm trained with stochastic gradient descent, a boosting algorithm comprising training data weighting, a boosting algorithm comprising updating training data weighting, or a boosting algorithm comprising updating misclassified training data with higher weights.
18 . The method of claim 1 , further comprising providing a pre-treatment image to the user as a guide for identifying the one or more of (a) the at least one anatomical feature or (b) the at least one anatomical boundary on the real-time operating field image.
19 . The method of claim 18 , wherein the pre-treatment image comprises one or more of an MM image, an X-ray image, a CT image, or an ultrasound image.
20 . A computer-implemented method of planning and implementing a treatment procedure, the method comprising:
(i) displaying a real-time operating field image to a user;
(ii) identifying one or more of (a) at least one anatomical feature or (b) at least one anatomical boundary on the real-time operating field image, with the aid of the trained classifier model, and without the aid of any label used to identify the at least one anatomical feature and without the aid of any manually made marking by the user for the at least one anatomical boundary;
(iii) displaying one or more of (a) at least one label for the identified at least one anatomical feature or (b) at least one marking for the identified at least one anatomical boundary on the displayed real-time operating field image;
(iv) allowing the user to modify the real-time operating field image by one or more of (a) re-labelling the identified at least one anatomical feature or (b) re-positioning the at least one marking for the at least one anatomical boundary; and
(v) providing an update to the trained classifier model based on the modifications by the user.
21 . The method of claim 20 , further comprising prompting the user to confirm that the at least one label for the identified at least one anatomical feature or (b) the at least one marking for the identified at least one anatomical boundary on the displayed real-time operating field image is correct, wherein step (iv) is performed if the user does not confirm that the at least one label for the identified at least one anatomical feature or (b) the at least one marking for the identified at least one anatomical boundary on the displayed real-time operating field image is correct.
22 . The method of claim 3 , further comprising treating an anatomical structure based on one or more of (a) the displayed at least one label for the at least one identified anatomical feature or (b) the displayed at least one marking for the at least one anatomical boundary on the real-time operating field image.