Systems and methods for human-machine partnered ptsd prediction

Abstract

Provided is a method for predicting a PTSD diagnosis in a patient comprising receiving audio input data from a patient; determining one or more audio input indicators based on the audio input data, wherein each audio input indicator of the one or more audio input indicators represents a likelihood of a positive PTSD diagnosis based on the audio input data; receiving clinical assessment data from the patient; determining one or more clinical assessment indicators based on the clinical assessment data, wherein each clinical assessment indicator of the one or more clinical assessment indicators represents a likelihood of a positive PTSD diagnosis based on the clinical assessment data; combining the one or more audio input indicators and the one or more clinical assessment indicators using a prediction model chosen by a clinician; and determining a PTSD diagnosis in the patient based on the audio input data and the clinical assessment data.

1 . A method for determining a PTSD likelihood in a patient comprising:
 receiving, at an electronic device, audio input data from a patient;   determining one or more audio input indicators based on the audio input data, wherein each audio input indicator of the one or more audio input indicators represents a likelihood of a positive PTSD diagnosis based on the audio input data;   receiving, at the electronic device, clinical assessment data from the patient, wherein the clinical assessment data comprises one or more of social support data, suicide ideation and attempts data, depression severity data, or self-assessment data;   determining one or more clinical assessment indicators based on the clinical assessment data, wherein each clinical assessment indicator of the one or more clinical assessment indicators represents a likelihood of a positive PTSD diagnosis based on the clinical assessment data;   combining the one or more audio input indicators and the one or more clinical assessment indicators using at least one machine learning model; and   determining, by the at least one machine learning model, the PTSD diagnosis likelihood in the patient based on the audio input data and the clinical assessment data.   
     
     
         2 . The method of  claim 1 , wherein the at least one machine learning model comprises an ensemble model. 
     
     
         3 . The method of  claim 2 , wherein the ensemble model determines the PTSD diagnosis likelihood in the patient based on the audio input data and the clinical assessment data using a plurality of machine learning algorithms. 
     
     
         4 . The method of  claim 1 , wherein the at least one machine learning model comprises any one or more of a regression model, a decision tree model, a neural network model, a support vector machine model, a random forest model, and a weighted ensemble model. 
     
     
         5 . The method of  claim 1 , wherein each audio input indicator of the one or more audio input indicators represents speech emotion, vocal features, or lexical features of the audio input data from the patient. 
     
     
         6 . The method of  claim 5 , wherein a first audio input indicator of the one or more audio input indicators comprises a speech emotion indicator, wherein the speech emotion indicator represents a likelihood of PTSD based on speech emotion extracted from the audio input data from the patient and compared to speech emotion training data at a machine-learned model trained on speech emotion data consistent with a positive PTSD diagnosis and speech emotion data consistent with a negative PTSD diagnosis. 
     
     
         7 . The method of  claim 5 , wherein a second audio input indicator of the one or more audio input indicators comprises a vocal features indicator, wherein the vocal features indicator represents a likelihood of PTSD based on vocal features extracted from the audio input data from the patient and compared to vocal features training data at a machine-learned model trained on vocal feature data consistent with a positive PTSD diagnosis and vocal feature data consistent with a negative PTSD diagnosis. 
     
     
         8 . The method of  5 , wherein a third audio input indicator of the one or more audio input indicators comprises a lexical features indicator, wherein the lexical features indicator represents a likelihood of a positive PTSD diagnosis based on lexical features extracted from the audio input data from the patient and compared to lexical features training data at a machine-learned model trained on lexical feature data consistent with a positive PTSD diagnosis and lexical feature data consistent with a negative PTSD diagnosis. 
     
     
         9 . The method of  claim 1 , wherein each clinical assessment indicator of the one or more clinical assessment indicators represents social support, suicide ideation and attempts, depression severity, or self-assessment of the clinical assessment data. 
     
     
         10 . The method of  claim 9 , wherein a first clinical assessment indicator of the one or more clinical assessment indicators comprises a social support indicator, wherein the social support indicator represents a likelihood of a positive PTSD diagnosis based on social support features extracted from the audio input data from the patient and compared to social support training data at a machine-learned model trained on social support data consistent with a positive PTSD diagnosis and social support data consistent with a negative PTSD diagnosis. 
     
     
         11 . The method of  claim 9 , wherein a second clinical assessment indicator of the one or more clinical assessment indicators comprises a suicide ideation and attempts indicator, wherein the suicide ideation and attempts indicator represents a likelihood of a positive PTSD diagnosis based on suicide ideation and attempts features extracted from the audio input data from the patient and compared to suicide ideation and attempts training data at a machine-learned model trained on suicide ideation and attempts data consistent with a positive PTSD diagnosis and suicide ideation and attempts data consistent with a negative PTSD diagnosis. 
     
     
         12 . The method of  claim 9 , wherein a third clinical assessment indicator of the one or more clinical assessment indicators comprises a depression severity indicator, wherein the depression severity indicator represents a likelihood of a positive PTSD diagnosis based on depression severity features extracted from the audio input data from the patient and compared to depression severity training data at a machine-learned model trained on depression severity data consistent with a positive PTSD diagnosis and depression severity data consistent with a negative PTSD diagnosis. 
     
     
         13 . The method of  claim 1 , wherein combining the one or more audio input indicators and the one or more clinical assessment indicators comprises the clinician adjusting the weight of one or more of a speech emotion indicator, a vocal features indicator, a lexical features indicator, a social support indicator, a suicide ideation and attempts indicator, a depression severity indicator, or a self-assessment indicator. 
     
     
         14 . A system for determining a PTSD likelihood in a patient comprising:
 a memory;   one or more processors; and   one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs when executed by the one or more processors cause the processor to:
 receive, at an electronic device, audio input data from a patient; 
 determine one or more audio input indicators based on the audio input data, wherein each audio input indicator of the one or more audio input indicators represents a likelihood of a positive PTSD diagnosis based on the audio input data; 
 receive, at the electronic device, clinical assessment data from the patient, wherein the clinical assessment data comprises one or more of social support data, suicide ideation and attempts data, depression severity data, or self-assessment data; 
 determine one or more clinical assessment indicators based on the clinical assessment data, wherein each clinical assessment indicator of the one or more clinical assessment indicators represents a likelihood of a positive PTSD diagnosis based on the clinical assessment data; 
 combine the one or more audio input indicators and the one or more clinical assessment indicators using at least one machine learning model; and 
 determine, by the at least one machine learning model, the PTSD diagnosis likelihood in the patient based on the audio input data and the clinical assessment data. 
   
     
     
         15 . A non-transitory computer readable storage medium storing one or more programs for determining a PTSD likelihood in a patient, the one or more programs comprising instructions, which, when executed by an electronic device with a display and a user input interface, cause the device to:
 receive, at an electronic device, audio input data from a patient;   determine one or more audio input indicators based on the audio input data, wherein each audio input indicator of the one or more audio input indicators represents a likelihood of a positive PTSD diagnosis based on the audio input data;   receive, at the electronic device, clinical assessment data from the patient, wherein the clinical assessment data comprises one or more of social support data, suicide ideation and attempts data, depression severity data, or self-assessment data;   determine one or more clinical assessment indicators based on the clinical assessment data, wherein each clinical assessment indicator of the one or more clinical assessment indicators represents a likelihood of a positive PTSD diagnosis based on the clinical assessment data;   combine the one or more audio input indicators and the one or more clinical assessment indicators using at least one machine learning model; and   determine, by the at least one machine learning model, the PTSD diagnosis likelihood in the patient based on the audio input data and the clinical assessment data.