Predictive visualization for aesthetic medical procedures

Abstract

A method includes training a machine learning model to generate predicted images to obtain a trained machine learning model, based on: a) pre-treatment training images; b) a plan of treatment; and c) post-treatment training images; where the plan of treatment includes: a) a first mark identifying where to apply a product, b) a first product to be applied at the first mark, and c) a first volume of the first product to be applied at the first mark; generating a predicted post-treatment image by applying the trained predictive visualization machine learning model to a new pre-treatment image, based on: a) a second mark on a new pre-treatment image of the area of a patient, b) a second product to be applied at the second mark, and c) a second volume of the second product to be applied at the second mark; where the predicted images identifies a modified area.

What is claimed is: 
     
         1 . A method comprising:
 training, by a processor, a predictive visualization machine learning model to generate predicted post-treatment images to obtain a trained predictive visualization machine learning model, based at least in part on:
 a) a set of pre-treatment training images of at least one area of a human; 
 b) a plan of treatment related to the set of pre-treatment training images; and 
 c) a set of post-treatment training images of the at least one area of the human related to the set of pre-treatment training images and the plan of treatment;
 wherein the plan of treatment comprises:
 a) at least one first treatment mark identifying where a product is to be applied on a pre-treatment image, 
 b) a first product to be applied at the at least one first treatment mark, and 
 c) a first volume of the product to be applied at the at least one first treatment mark; 
 
 
   generating, by the processor, at least one predicted post-treatment image by applying the trained predictive visualization machine learning model to at least one new pre-treatment image, based at least in part on a new plan of treatment comprising:
 a) at least one second treatment mark on a new pre-treatment image of the at least one area of a patient, 
 b) a second product to be applied at the at least one second treatment mark, and 
 c) a second volume of the second product to be applied at the at least one second treatment mark; 
 wherein the at least one predicted post-treatment image identifies at least one modified area; and 
   instructing, by the processor, to display the at least one predicted post-treatment image on a screen.   
     
     
         2 . The method of  claim 1 , further comprising:
 receiving, by the processor, from the patient:
 a) at least one patient image; and 
 b) at least one patient treatment request; 
   wherein the new plan of treatment is based at least in part on the at least one patient image and the at least one patient treatment request.   
     
     
         3 . The method of  claim 1 , wherein the predictive visualization machine learning model includes one or more of a neural network, a radial basis function network, an image classifier, a recurrent neural network, a convolutional network, a generative adversarial network, a fully connected neural network, a feedforward neural network, or a combination thereof. 
     
     
         4 . The method of  claim 1 , wherein the predictive visualization machine learning model applies at least one loss function to the set of post-treatment training images. 
     
     
         5 . The method of  claim 4 , wherein the at least one loss function comprises a mean square error loss function, an internal adversarial network, an opensource adversarial network, or a combination thereof. 
     
     
         6 . The method of  claim 1 , wherein the first product comprises at least one of a prescription injection or a dermal filler. 
     
     
         7 . The method of  claim 1 , wherein the second product and the first product are the same. 
     
     
         8 . The method of  claim 1 , wherein the predictive visualization machine learning model is trained on thousands of pre-treatment training images and post-treatment training images. 
     
     
         9 . The method of  claim 1 , further comprising applying, by the processor, a registration process to finetune an alignment of the set of pre-treatment training images with an alignment of the set of post-treatment training images. 
     
     
         10 . The method of  claim 9 , wherein the registration process identifies from 10 to 500 facial landmarks on the set of pre-treatment training images and the set of post-treatment training images.