Content based image retrieval for lesion analysis
Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) are commonly used to assess patients with known or suspected pathologies of the lungs and liver. In particular, identification and quantification of possibly malignant regions identified in these high-resolution images is essential for accurate and timely diagnosis. However, careful quantitative assessment of lung and liver lesions is tedious and time consuming. This disclosure describes an automated end-to-end pipeline for accurate lesion detection and segmentation.
1. A machine learning system, comprising:
at least one nontransitory processor-readable storage medium that stores at least one of processor-executable instructions or data; and
at least one processor communicably coupled to the at least one nontransitory processor-readable storage medium, in operation the at least one processor:
loads a trained machine learning model that has been designed to predict patients' clinical outcomes, wherein the machine learning model is trained based at least partially on medical images that indicate longitudinal changes of suspected lesions over time;
calculates features for a query patient who has a known or suspected diagnosis, based at least partially on medical images of the query patient that indicate longitudinal changes of suspected lesions over time;
identifies a selection of treatment options for the query patient from a list of available treatment options; and
for each of one or more of the identified treatment options, uses the features and treatment options as inputs to the trained machine learning model to generate prediction results, the prediction results comprising a prediction of the clinical outcomes for the query patient.
2. The machine learning system of claim 1 wherein, in operation, the at least one processor presents the prediction results of the machine learning model to a user via a display.
3. The machine learning system of claim 2 wherein, in operation, the at least one processor presents the prediction results in tabular format.
4. The machine learning system of claim 3 wherein the treatment options form one axis of the table and at least some of the cells of the table indicate the likelihood of occurrence of a given clinical outcome.
5. The machine learning system of claim 3 wherein at least some of the cells of the table indicate lower and upper confidence interval boundaries of the likelihood of occurrence of a given clinical outcome.
6. The machine learning system of claim 2 wherein, in operation, the at least one processor presents the prediction results in the form of a chart.
7. The machine learning system of claim 6 wherein one axis of the chart shows different treatment options and the other axis of the chart indicates likelihood of occurrence of a given clinical outcome.
8. The machine learning system of claim 6 wherein the chart format is a bar chart.
9. The machine learning system of claim 6 wherein the chart indicates upper and lower confidence interval boundaries of the prediction results.
10. The machine learning system of claim 1 wherein at least one treatment option is a combination of different treatments.
11. The machine learning system of claim 1 wherein the at least one processor permits a user to select the selection of treatment options.
12. The machine learning system of claim 1 wherein at least some of the features calculated by the at least one processor are clinical features.
13. The machine learning system of claim 12 wherein the clinical features include one or more of patient demographic information, patient medical history, family medical history or diagnostic information related to a current lesion.
14. The machine learning system of claim 1 wherein at some of the features calculated by the at least one processor are calculated from medical images of the query patient.
15. The machine learning system of claim 14 wherein at least some of the calculated features are calculated using one or more pre-trained CNN models.
16. The machine learning system of claim 15 wherein the one or more pre-trained CNN models include at least one of a classification model, an object detection model or a semantic segmentation model.
17. The machine learning system of claim 1 wherein at least some of the clinical outcomes predicted by the at least one processor include one or more of cancer-related mortality, overall mortality, response to treatment, cancer recurrence, medical complications, adverse events, patient quality of life or optimal treatment.
18. The machine learning system of claim 1 wherein the trained machine learning model that is used by the at least one processor to predict the query patient's clinical outcomes is an ensemble of one or more of a random forest, gradient boosted decision trees or a multi-layer perceptron.
19. The machine learning system of claim 1 wherein the query patient has known or suspected lung cancer.
20. The machine learning system of claim 1 wherein the query patient has known or suspected liver cancer.
21. A computer-implemented method, comprising:
loading a trained machine learning model that has been designed to predict patients' clinical outcomes, wherein the machine learning model is trained based at least partially on medical images that indicate longitudinal changes of suspected lesions over time;
calculating features for a query patient who has a known or suspected diagnosis, based at least partially on medical images of the query patient that indicate longitudinal changes of suspected lesions over time;
identifying a selection of treatment options for the query patient from a list of available treatment options; and
for each of one or more of the identified treatment options, using the features and treatment options as inputs to the trained machine learning model to generate prediction results, the prediction results comprising a prediction of the clinical outcomes for the query patient.
22. The method of claim 21 wherein the prediction results further indicate at least one of a likelihood of occurrence of a given clinical outcome or different treatment options.
23. The method of claim 21 wherein at least some of the features calculated include one or more of patient demographic information, patient medical history, family medical history or diagnostic information related to a current lesion.
24. The method of claim 21 wherein at least some of the features are calculated based, at least in part, on medical images of the query patient.
25. The method of claim 24 wherein at least some of the features are calculated using one or more pre-trained convolutional neural network (CNN) models.
26. The method of claim 25 wherein the one or more pre-trained CNN models include at least one of a classification model, an object detection model or a semantic segmentation model.
27. The method of claim 21 wherein at least some of the clinical outcomes predicted include one or more of cancer-related mortality, overall mortality, response to treatment, cancer recurrence, medical complications, adverse events, patient quality of life or optimal treatment.
28. The method of claim 21 wherein the trained machine learning model includes an ensemble of one or more of a random forest, gradient boosted decision trees or a multi-layer perceptron.