Decision support system for medical therapy planning
For decision support in a medical therapy, machine learning provides a machine-learned generator for generating a prediction of outcome for therapy personalized to a patient. Deep learning may result in features more predictive of outcome than handcrafted features. More comprehensive learning may be provided by using multi-task learning where one of the tasks (e.g., segmentation, non-image data, and/or feature extraction) is unsupervised and/or draws on a greater number of training samples than available for outcome prediction alone.
1. A method for decision support in a medical therapy system, the method comprising:
acquiring a medical scan of a patient;
generating a prediction of outcome of therapy for the patient from scan data of the medical scan, the outcome generated by a machine-learned multi-task generator having been trained with unsupervised learning from training data unlabeled for the outcome and having been trained from labeled training data for the outcome; and
displaying an image of the outcome.
2. The method of claim 1 wherein acquiring comprises scanning the patient with a computed tomography scanner.
3. The method of claim 1 wherein acquiring comprises acquiring voxel data representing a three-dimensional distribution of locations in a volume of the patient, and wherein generating comprises generating based on input of the voxel data.
4. The method of claim 1 wherein generating comprises generating with the machine-learned multi-task generator comprises an autoencoder having been trained with the unsupervised learning.
5. The method of claim 4 wherein generating comprises generating with the training data unlabeled for the outcome comprising tumor segmentation, non-image data, and/or radiomic image features.
6. The method of claim 4 wherein generating comprises generating with the machine-learned multi-task generator comprising a fully-connected network configured to receive input from a bottleneck of the autoencoder, the fully-connected network having been trained from the labeled training data.
7. The method of claim 6 wherein generating comprises generating in response to input of a scale and intensity normalized version of the scan data and in response to input of a scale, intensity maximum, and intensity minimum for the version of the medical scan.
8. The method of claim 4 wherein generating comprises generating with the autoencoder comprising an encoder network trained as part of an encoder and decoder network and with the machine-learned multi-task generator comprising a neural network configured to receive bottleneck features of the encoder network, the neural network generating the outcome.
9. The method of claim 1 further comprising applying therapy to the patient based on the outcome.
10. The method of claim 1 further comprising clustering image features from a bottleneck of an encoder of the machine-learned multi-task generator and identifying another patient based on the clustering.
11. A medical imaging system for therapy decision support, the medical imaging system comprising:
a medical imager configured to scan a patient;
an image processor configured to predict a result of therapy for the patient in response to input of scan data from the scan to a machine-trained network, the machine-trained network including an encoder of an autoencoder trained in an unsupervised manner and a fully-connected network configured to receive an output of the encoder to predict the result; and
a display configured to display the predicted result.
12. The medical imaging system of claim 11 wherein the medical imager comprises a computed tomography imager, and wherein the machine-trained network was trained using a first loss for image features or segmentation and using a second loss for outcome.
13. The medical imaging system of claim 11 wherein the machine-trained network comprises the encoder where the autoencoder comprised the encoder and a decoder to estimate an input from the output of the encoder in the training in the unsupervised manner.
14. The medical imaging system of claim 11 wherein the fully-connected network comprises a fully-connected multi-layer perceptron.
15. The medical imaging system of claim 11 wherein the image processor is configured to cluster the output of the encoder and identify another patient based on the cluster of the output of the encoder.
16. A method for decision support in a medical therapy system, the method comprising:
acquiring a medical scan of a patient;
generating values for bottleneck features of a machine-learned autoencoder having been trained with unsupervised learning from training data unlabeled for a therapy outcome;
clustering the values for the bottleneck features;
identifying another patient based on the values as clustered; and
displaying therapy data for the other patient.
17. The method of claim 16 wherein acquiring comprises scanning the patient with a computed tomography scanner.
18. The method of claim 16 wherein generating further comprises generating with the machine-learned autoencoder comprising a multi-task generator configured to generate an output of therapy for the patient from the values for the bottleneck features.