Machine learning-assisted computed tomography workflow
A system and method include acquisition of a two-dimensional radiograph of a patient, input of the radiograph to a trained machine learning model to generate a classification, performance of a computed tomography scan of the patient based on the two-dimensional radiograph if the classification indicates that the patient does not have a first condition, and determination to modify the computed tomography scan if the classification indicates that the patient has the first condition.
1. A system comprising:
an imaging system; and
a computing system to:
plan a computed tomography scan of a patient;
operate the imaging system to acquire a two-dimensional radiograph of the patient;
input the radiograph to a trained machine learning model;
receive from the trained machine learning model, in response to the input radiograph, a classification indicating whether the patient is pregnant or not pregnant;
if the classification indicates that the patient is not pregnant, operate the imaging system to perform the computed tomography scan of the patient based on the two-dimensional radiograph; and
if the classification indicates that the patient is pregnant, shield a fetus within the patient, modify the computed tomography scan to reduce radiation delivered to reproductive organs of the patient, and operate the imaging system to perform the modified computed tomography scan.
2. A system according to claim 1 , wherein the model is trained based on a plurality of radiographs and a classification corresponding to each radiograph, wherein the classification corresponding to a radiograph is determined based on a computed tomography volume generated contemporaneously to the radiograph.
3. A system according to claim 1 ,
wherein the model is trained based on a plurality of radiographs and a plurality of classifications corresponding to each radiograph, wherein the plurality of classifications corresponding to a radiograph are determined based on a computed tomography volume generated contemporaneously to the radiograph.
4. A method comprising:
planning a computed tomography scan of a patient;
acquiring a two-dimensional radiograph of the patient;
inputting the radiograph to a trained machine learning model;
receiving from the trained machine learning model, in response to the input radiograph, a classification indicating whether the patient is pregnant or not pregnant;
if the classification indicates that the patient is not pregnant, performing the computed tomography scan of the patient based on the two-dimensional radiograph; and
if the classification indicates that the patient is pregnant, shielding a fetus within the patient, modifying the computed tomography scan to reduce radiation delivered to reproductive organs of the patient, and operating an imaging system to perform the modified computed tomography scan.
5. A method according to claim 4 , wherein the model is trained based on a plurality of radiographs and a classification corresponding to each radiograph, wherein the classification corresponding to a radiograph is determined based on a computed tomography volume generated contemporaneously to the radiograph.
6. A method according to claim 4 ,
wherein the model is trained based on a plurality of radiographs and a plurality of classifications corresponding to each radiograph, wherein the plurality of classifications corresponding to a radiograph are determined based on a computed tomography volume generated contemporaneously to the radiograph.
7. A system comprising:
a storage device storing a trained classification model; and
a processing unit to:
plan a computed tomography scan of a patient;
input a two-dimensional radiograph of the patient to the trained classification model;
receive from the trained machine learning model, in response to the input radiograph, a classification indicating whether the patient is pregnant or not pregnant;
if the classification indicates that the patient is not pregnant, operate an imaging system to perform the computed tomography scan of the patient based on the two-dimensional radiograph; and
if the classification indicates that the patient is pregnant, shield a fetus within the patient, modify the computed tomography scan to reduce radiation delivered to reproductive organs of the patient, and operate the imaging system to perform the modified computed tomography scan.
8. A system according to claim 7 , wherein the model is trained based on a plurality of radiographs and a classification corresponding to each radiograph, wherein the classification corresponding to a radiograph is determined based on a computed tomography volume generated contemporaneously to the radiograph.
9. A system according to claim 7 ,
wherein the model is trained based on a plurality of radiographs and a plurality of classifications corresponding to each radiograph, wherein the plurality of classifications corresponding to a radiograph are determined based on a computed tomography volume generated contemporaneously to the radiograph.