Incorporating clinical and economic objectives for medical AI deployment in clinical decision making
An AI algorithm may be used in a clinical setting to perform one or more tasks to assist medical personnel. The results produced by the AI algorithm may affect not only patient care, but also the cost of the care. The AI algorithm may be trained on auxiliary data to incorporate the impacts on patient care and cost.
1 . A method for assessing medical images, the method comprising:
receiving, by a processor, first medical image data and first auxiliary patient data;
identifying, by the processor, a plurality of machine-learned models tuned to various economic and clinical environments;
inputting, by the processor, the first auxiliary patient data to a machine-learned classifier, the machine-learned classifier trained on second auxiliary patient data, wherein the second auxiliary patient data comprises at least indicators of clinical impact and economic impact for a particular region and/or a healthcare facility of the results generated by the plurality of machine-learned models;
outputting, by the machine-learned classifier in response to inputting the first auxiliary patient data to the machine-learned classifier, a machine-learned model of the plurality of machine-learned models, wherein the machine-learned model or a performance specification for the machine-learned model is tuned to individual constraints of a particular environment described by the first auxiliary patient data;
applying, by the processor, the first medical image data and the first auxiliary patient data to the machine-learned model;
generating, by the processor, a medical image result of the first medical image data based on applying the first medical image data and the first auxiliary patient data to the machine-learned model; and
outputting, by the processor, the medical image result.
2 . The method of claim 1 , wherein the medical image result comprises a label, a segmentation, a classification, a registration, a quantification, a fusion image, a simulation, or a combination thereof.
3 . The method of claim 1 , wherein the machine-learned model is a detection algorithm, a segmentation algorithm, a quantification algorithm, a model, a registration algorithm, a fusion algorithm, a simulation algorithm, or a combination thereof.
4 . The method of claim 1 , wherein performance specification comprises an operating characteristic curve, an operating point, or a combination thereof.
5 . The method of claim 1 , wherein the first auxiliary patient data comprises patient vital information, patient outcome statistics, available treatment options, medical treatment guidelines, length of stay, treatment cost, reimbursement codes, mistakes made, or a combination thereof.
6 . The method of claim 1 , wherein the first auxiliary patient data is received from a network interface in communication with a medical data repository.
7 . The method of claim 1 , further comprising:
receiving, by the processor, further auxiliary patient data collected based on the medical image result; and
updating, by the processor, the machine-learned model based on applying the further auxiliary patient data to the machine-learned classifier.
8 . The method of claim 1 , wherein the machine-learned classifier and the machine-learning model are integrated into a machine learning network, and
wherein the first medical image data and first auxiliary patient data are applied to the machine learning network.
9 . A medical image assessment system, the system comprising:
an image processor, coupled with a memory containing instructions that, when executed, cause the image processor to:
apply first auxiliary patient data to a machine-learned classifier;
select, by the machine-learned classifier in response to applying the first auxiliary patient data, a machine-learned model from a plurality of machine-learned models, the selected machine-learned model tuned to individual constraints of a particular environment described by the first auxiliary patient data, wherein the machine-learned classifier is trained using at least second auxiliary patient data comprising at least indicators of clinical and economic impacts for a particular region and/or healthcare facility of the results generated by the plurality of machine-learned models;
apply a first medical image to the selected machine-learned model; and
generate an analysis of the first medical image based on the application of the first medical image to the machine-learned model.
10 . The system of claim 9 , wherein each machine-learned model of the plurality of machine-learned models is configured with a different operating characteristic curve of a plurality of operating characteristic curves.
11 . The system of claim 9 , wherein each machine-learned model of the plurality of machine-learned models is configured with a different operating point of a plurality of operating points.
12 . A method for training a machine-learning model, the method comprising:
receiving, by a processor, auxiliary patient data and a plurality of machine learning specifications associated with the auxiliary patient data that are tuned to specific environments related to the auxiliary patient data, wherein the auxiliary patient data comprises at least indicators of clinical and economic impact for a particular region and/or a healthcare facility;
training with machine learning, by the processor, the machine-learning model based on the auxiliary patient data and the machine learning specifications, wherein a result of the training is a machine-learned model that is configured to output a machine learning specification from the plurality of machine learning specifications that is tuned to individual constraints of a particular environment; and
storing, by the processor, the machine-learned model.
13 . The method of claim 12 , wherein the plurality of machine learning specifications comprises a plurality of operating characteristic curves.
14 . The method of claim 12 , wherein the plurality of machine learning specifications comprises a plurality of operating points.
15 . The method of claim 12 , wherein the plurality of machine learning specifications comprises a plurality of further machine-learned models.
16 . The method of claim 12 , further comprising:
applying, by the processor, second auxiliary patient data to the machine-learned model; and
generating, by the processor, the output based on the applying,
wherein the machine learning specification is based on the second auxiliary patient data.
17 . The method of claim 12 , further comprising:
receiving, by the processor, third auxiliary patient data collected based on the output; and
updating, by the processor, the machine-learned model based on applying the third auxiliary patient data and the plurality of machine learning specifications to the machine-learned model.