Bone suppression for chest radiographs using deep learning
A system and method for generating a rib suppressed radiographic image using deep learning computation. The method includes using a convolutional neural network module trained with pairs of a chest x-ray image and its counterpart bone suppressed image. The bone suppressed image is obtained using a bone suppression algorithm applied to the chest x-ray image. The convolutional neural network module is then applied to a chest x-ray image or the bone suppressed image to generate an enhanced bone suppressed image.
1. A system comprising:
an x-ray source;
an x-ray detector;
a processor programmed to capture only one x-ray image of a patient, using the x-ray detector and only one exposure of the x-ray source, and to apply a bone suppression algorithm on the one captured x-ray image of the patient to generate a bone suppressed version of the one captured x-ray image of the patient, thereby forming a first radiographic image pair consisting of the one captured x-ray image and the bone suppressed version of the one captured x-ray image; and
a deep learning module, wherein the processor is further programmed to train the deep learning module using the first radiographic image pair and a collection of additional radiographic image pairs each consisting of only one x-ray image captured with the detector using only one exposure of the x-ray source and a bone suppressed version thereof.
2. The system of claim 1 , wherein the processor is configured to apply the trained deep learning module to a newly captured radiographic image of a patient to generate an enhanced bone suppressed version of the newly captured radiographic image of the patient.
3. A computer implemented method comprising:
forming a plurality of radiographic image pairs each comprising a starting image and a target image, the plurality of radiographic image pairs each formed by the steps of:
capturing only one digital radiographic image of a patient to form the starting image; and
applying a bone suppression algorithm to the one captured digital radiographic image of the patient to form the target image; and
providing a convolutional neural network module and training the module using the plurality of radiographic image pairs.
4. The method of claim 3 , wherein the step of capturing includes capturing only one chest digital radiographic image, and wherein the step of applying includes forming a rib-suppressed target image.
5. The system of claim 1 , wherein the processor is programmed to capture only one new x-ray image of a patient, using the x-ray detector and only one exposure of the x-ray source, to apply a bone suppression algorithm on the one new captured x-ray image of the patient to generate a bone suppressed version thereof, and to apply the trained deep learning module to the bone suppressed version to generate an enhanced bone suppressed version of the new captured x-ray image of the patient.
6. The method of claim 3 , further comprising capturing a new digital x-ray image of a patient and applying the convolutional neural network module to the new digital x-ray image of a patient to generate an enhanced bone suppressed radiographic image of the patient.
7. The method of claim 3 , further comprising capturing a new digital x-ray image of a patient, applying a bone suppression algorithm to the new digital x-ray image of the patient to generate a bone-suppressed digital x-ray image of the patient, and applying the convolutional neural network module to the bone suppressed digital x-ray image of the patient to generate an enhanced bone suppressed radiographic image of the patient.