Systems and methods for producing a brain lesion functional MRI biomarker, predicting patient prognosis, and treatment planning
A biomarker predictive of a survival outcome of a brain tumor patient is disclosed. The biomarker includes a functional connectivity matrix that includes a plurality of matrix elements. Each matrix element includes a correlation of resting-state fMRI activities of a first and second region of interest from a plurality of regions of interest within the patient's brain. Computing device and systems are disclosed to transform a resting-state fMRI dataset obtained from the patient into the biomarker and to transform the biomarker into a predicted survival outcome using a machine learning model.
1 . A computing device configured to transform a resting-state fMRI dataset of a brain tumor patient into a predicted survival outcome, the computing device comprising a display device, at least one processor, and a non-volatile computer-readable media, the non-volatile computer-readable media containing instructions executable on the at least one processor to:
a. transform the resting-state fMRI dataset into a functional connectivity matrix comprising a plurality of matrix elements, matrix each element comprising a correlation of resting-state fMRI activities of a first and second region of interest from a plurality of regions of interest within the patient's brain;
b. transform the functional connectivity matrix into the predicted survival outcome using a machine learning model;
c. display the predicted survival outcome on the display device; and
d. provide a brain mapping using the same resting-state fMRI dataset used in providing the predicted survival outcome, wherein the brain mapping is adapted for use in treatment planning.
2 . The computing device of claim 1 , wherein the correlation of resting-state fMRI activities comprises a Pearson's correlation coefficient.
3 . The computing device of claim 1 , wherein the brain tumor patient comprises a glioblastoma multiforme patient.
4 . The computing device of claim 1 , wherein each matrix comprises a correlation strength of resting-state fMRI activities.
5 . The computing device of claim 4 , wherein the correlation strength comprises a Fisher's z transformed Pearson's correlation strength.
6 . The computing device of claim 1 , wherein the machine learning model is pre-trained using a training set comprising resting-state fMRI datasets from a plurality of glioblastoma multiforme patients and corresponding survival outcomes.
7 . The computing device of claim 1 , wherein the predicted survival outcome comprises a classification of the patient as a short-term survivor or a long-term survivor, wherein the short-term survivor indicates a predicted survival of less than a threshold survival term and the long-term survivor indicates a predicted survival of greater than the threshold survival term.
8 . The computing device of claim 1 , wherein the predicted survival outcome comprises a predicted survival duration of the patient.
9 . The computing device of claim 1 , wherein the machine learning model transforms the functional connectivity matrix into a predicted survival outcome using at least one of a classification analysis and a regression analysis.
10 . The computing device of claim 1 , wherein the machine learning model comprises one of a deep convolutional neural network and a support machine vector with a linear kernel.
11 . The computing device of claim 1 , wherein the non-volatile computer-readable media further contains instructions executable on the at least one processor to recommend a treatment based on the predicted survival outcome.
12 . The computing device of claim 1 , wherein the treatment for which the brain mapping is adapted for planning comprises resection of a brain tumor.
13 . The computing device of claim 1 , wherein the treatment for which the brain mapping is adapted for planning comprises at least one of radiation treatment and chemotherapy.
14 . The computing device of claim 1 , wherein the brain mapping comprises mappings of a set of canonical resting state network (“RSNs”) generated using the machine learning model.
15 . The computing device of claim 14 , wherein the set of canonical RSNs includes at least a sensorimotor RSN (“SMN”), a language RSN (“LAN”), and a vision RSN (“VIS”).
16 . The computing device of claim 14 , wherein transforming the functional connectivity matrix into the predicted survival outcome uses the mappings of the set of canonical RSNs generated using the machine learning model.
17 . The computing device of claim 16 , wherein transforming the functional connectivity matrix into the predicted survival outcome further comprises determining a boundary of the brain tumor.
18 . The computing device of claim 17 , wherein transforming the functional connectivity matrix into the predicted survival outcome further comprises evaluating intratumoral resting state functional connectivity.
19 . The computing device of claim 18 , wherein evaluating intratumoral resting state functional connectivity comprises evaluating functional connectivity of brain tissue residing within the tumor boundary with at least one of the canonical RSNs.
20 . The computing device of claim 1 , wherein transforming the functional connectivity matrix into the predicted survival outcome further comprises determining a boundary of the brain tumor.
21 . The computing device of claim 20 , wherein transforming the functional connectivity matrix into the predicted survival outcome further comprises evaluating intratumoral resting state functional connectivity.
22 . A computer-implemented method of predicting a survival outcome of a brain tumor patient, the method comprising:
a. receiving, at a computing device, a resting-state fMRI dataset of the brain tumor patient;
b. transforming, using the computing device, the resting-state fMRI dataset into a functional connectivity matrix comprising a plurality of matrix elements, each matrix element comprising a correlation of resting-state fMRI activities of a first and second region of interest from a plurality of regions of interest within the patient's brain;
c. transforming, using the computing device, the functional connectivity matrix into the predicted survival outcome using a machine learning model;
d. displaying the predicted survival outcome on a display device; and
e. providing a brain mapping using the same resting-state fMRI dataset used in providing the predicted survival outcome, wherein the brain mapping is adapted for use in treatment planning.
23 . The computer-implemented method of claim 22 , wherein the correlation of resting-state fMRI activities comprises a Pearson's correlation coefficient.
24 . The computer-implemented method of claim 22 , wherein the brain tumor patient comprises a glioblastoma multiforme patient.
25 . The computer-implemented method of claim 22 , wherein each matrix element comprises a correlation strength of resting-state fMRI activities.
26 . The computer-implemented method of claim 25 , wherein the correlation strength comprises a Fisher's z transformed Pearson's correlation strength.
27 . The computer-implemented method of claim 22 , wherein the machine learning model is pre-trained using a training set comprising resting-state fMRI datasets from a plurality of glioblastoma multiforme patients and corresponding survival outcomes.
28 . The computer-implemented method of claim 22 , wherein the predicted survival outcome comprises a classification of the patient as a short-term survivor or a long term survivor, wherein the short-term survivor indicates a predicted of survival less than a threshold survival term and the long-term survivor indicates a predicted survival of greater than the threshold survival term.
29 . The computer-implemented method of claim 22 , wherein the predicted survival outcome comprises a predicted survival duration of the patient.
30 . The computer-implemented method of claim 22 , wherein the machine learning model transforms the functional connectivity matrix into a predicted survival outcome using at least one of a classification analysis and a regression analysis.
31 . The computer-implemented method of claim 22 , wherein the machine learning model comprises one of a deep convolutional neural network and a support machine vector with a linear kernel.
32 . The computer-implemented method of claim 22 , further comprising recommending a treatment based on the predicted survival outcome.