Periodic Reporting for period 3 - BREAST4D (4D Breast Imaging for Personalized Breast Cancer Treatment)
Reporting period: 2023-06-01 to 2024-11-30
Yearly, over 130,000 women in Europe, and half a million worldwide, die from breast cancer, even though screening and improvements in treatment have halved breast cancer mortality.
After breast cancer diagnosis, in general, treatment planning is based on the already acquired diagnostic images. Advanced methods, like magnetic resonance imaging (MRI), is only sometimes used for assessing tumor size and detecting additional tumors in the breasts immediately after diagnosis. At some institutions, MRI is also used midway and at the end of chemotherapy performed before surgery. This chemotherapy is used about one quarter of the cases before surgery to shrink the tumor and allow for surgery which does not need the removal of the whole breast. Currently this chemotherapy is planned based on what is discovered about the tumor from the biopsy. However, tumors are known to have regions that are different within them, resulting in regions that react differently to different drugs. This variability in chemotherapy response may be due to this tumor heterogeneity, which is not considered during treatment planning. As a result, less than half of the chemotherapy treatments result in the entire tumor disappearing. Therefore, there is a need to develop a method to identify all the different characteristics of every tumor. This would maximize the chances of achieving complete response from the tumor during pre-surgery chemotherapy, improving the chances of surviving breast cancer, especially in patients with the most aggressive tumors.
To achieve this, an imaging modality that can see what the tumor is made of, how it behaves and works internally, with smaller details and faster processes is needed. Therefore, in BREAST4D, we will develop a new breast imaging method, called 4D breast computed tomography (CT), based on the same concepts of body CT, which is now common in every radiology department, and 3D breast CT, a new imaging method to find breast cancer and determine the shape and size of breast tumors. However, 3D breast CT cannot determine what tumors are made of or how they function. In BREAST4D, we will develop the hardware and software necessary to create 4D breast CT, and perform the initial patient tests to show that it could be used to improve treatment planning and monitoring of response to treatment.
An imaging method such as 4D breast CT would result in a better understanding of the tumor nature and environment, allowing for treatment planning to be optimized, increasing the chances of achieving complete removal of the tumor. In addition, by repeating the imaging during chemotherapy, any changes in response or the fact that the tumor disappeared before the end of the planned treatment, could be seen and then the treatment could be modified, as needed. Finally, if an imaging method were able to predict or detect that the tumor completely disappeared during chemotherapy, then it would be possible to avoid any breast surgery completely. This high-risk clinical development has the potential largest pay-off: the avoidance of surgery on all patients that achieved complete chemotherapy response. In cases in which complete disappearance of the tumor has not been achieved, the information obtained from 4D breast CT would provide exquisite information for surgery planning, maximizing the chances that the entire tumor is removed while minimizing the amount of healthy tissue removed.
Therefore, 4D DCE-BCT may impact breast cancer treatment in all its phases: staging, treatment planning, response prediction and monitoring, and final outcome determination. With its unparalleled combined high spatio-temporal resolution, this imaging modality has the potential to allow for a precision medicine approach to breast cancer treatment, reducing mortality while minimizing morbidity.
After breast cancer diagnosis, in general, treatment planning is based on the already acquired diagnostic images. Advanced methods, like magnetic resonance imaging (MRI), is only sometimes used for assessing tumor size and detecting additional tumors in the breasts immediately after diagnosis. At some institutions, MRI is also used midway and at the end of chemotherapy performed before surgery. This chemotherapy is used about one quarter of the cases before surgery to shrink the tumor and allow for surgery which does not need the removal of the whole breast. Currently this chemotherapy is planned based on what is discovered about the tumor from the biopsy. However, tumors are known to have regions that are different within them, resulting in regions that react differently to different drugs. This variability in chemotherapy response may be due to this tumor heterogeneity, which is not considered during treatment planning. As a result, less than half of the chemotherapy treatments result in the entire tumor disappearing. Therefore, there is a need to develop a method to identify all the different characteristics of every tumor. This would maximize the chances of achieving complete response from the tumor during pre-surgery chemotherapy, improving the chances of surviving breast cancer, especially in patients with the most aggressive tumors.
To achieve this, an imaging modality that can see what the tumor is made of, how it behaves and works internally, with smaller details and faster processes is needed. Therefore, in BREAST4D, we will develop a new breast imaging method, called 4D breast computed tomography (CT), based on the same concepts of body CT, which is now common in every radiology department, and 3D breast CT, a new imaging method to find breast cancer and determine the shape and size of breast tumors. However, 3D breast CT cannot determine what tumors are made of or how they function. In BREAST4D, we will develop the hardware and software necessary to create 4D breast CT, and perform the initial patient tests to show that it could be used to improve treatment planning and monitoring of response to treatment.
An imaging method such as 4D breast CT would result in a better understanding of the tumor nature and environment, allowing for treatment planning to be optimized, increasing the chances of achieving complete removal of the tumor. In addition, by repeating the imaging during chemotherapy, any changes in response or the fact that the tumor disappeared before the end of the planned treatment, could be seen and then the treatment could be modified, as needed. Finally, if an imaging method were able to predict or detect that the tumor completely disappeared during chemotherapy, then it would be possible to avoid any breast surgery completely. This high-risk clinical development has the potential largest pay-off: the avoidance of surgery on all patients that achieved complete chemotherapy response. In cases in which complete disappearance of the tumor has not been achieved, the information obtained from 4D breast CT would provide exquisite information for surgery planning, maximizing the chances that the entire tumor is removed while minimizing the amount of healthy tissue removed.
Therefore, 4D DCE-BCT may impact breast cancer treatment in all its phases: staging, treatment planning, response prediction and monitoring, and final outcome determination. With its unparalleled combined high spatio-temporal resolution, this imaging modality has the potential to allow for a precision medicine approach to breast cancer treatment, reducing mortality while minimizing morbidity.
The 4D breast CT system was designed and built and is now installed at our laboratory.
A deep learning model was developed for x-ray scatter correction. This model was able to estimate scatter with high accuracy, resulting in scatter-corrected images with improved contrast and without significantly affecting contrast-to-noise ratio.
We are currently working on the construction of physical dynamic breast phantoms with programmable and variable contrast flow. Such phantoms will be used to test and optimize the entire imaging and analysis process more accurately. For this, we are developing a phantom that will contain different spherical inserts, representing tumors, containing inlet and outlet tubes. These tubes will allow for the flow of liquid, representing blood, containing iodinated contrast agent at different and varying concentrations. For this, the inlet will be connected to a programmable pump that allows to modify the flow rate and the source of the injected liquid. This phantom will eventually be developed further to represent realistic breast by incorporating a more advanced glandular/adipose texture and vascular network.
In parallel, a data-driven motion correction algorithm to compensate for non-rigid motion of the breast during acquisition of 3D and 4D breast CT imaging was developed. This method can correct for both sudden motion due to patient shifts and for periodic motion due to repetitive effects, such as breathing or the heartbeat. Since this is a purely data-driven process, the compensation method can remove motion artifacts for unknown but present motion patterns. As part of this research, it was discovered that image blurring due to motion is present extensively in 3D breast CT images, even without it being apparent to breast radiologists. The developed method can remove this motion artifact, improving the image quality in a substantial portion of cases.
Finally, a reconstruction method to quantify and reconstruct the iodine concentration throughout the imaged breast was implemented and initially tested. This method seems to recover the iodine content accurately in basic simulations, and is now being extended to be able to reconstruct 4D breast CT iodine images.
A deep learning model was developed for x-ray scatter correction. This model was able to estimate scatter with high accuracy, resulting in scatter-corrected images with improved contrast and without significantly affecting contrast-to-noise ratio.
We are currently working on the construction of physical dynamic breast phantoms with programmable and variable contrast flow. Such phantoms will be used to test and optimize the entire imaging and analysis process more accurately. For this, we are developing a phantom that will contain different spherical inserts, representing tumors, containing inlet and outlet tubes. These tubes will allow for the flow of liquid, representing blood, containing iodinated contrast agent at different and varying concentrations. For this, the inlet will be connected to a programmable pump that allows to modify the flow rate and the source of the injected liquid. This phantom will eventually be developed further to represent realistic breast by incorporating a more advanced glandular/adipose texture and vascular network.
In parallel, a data-driven motion correction algorithm to compensate for non-rigid motion of the breast during acquisition of 3D and 4D breast CT imaging was developed. This method can correct for both sudden motion due to patient shifts and for periodic motion due to repetitive effects, such as breathing or the heartbeat. Since this is a purely data-driven process, the compensation method can remove motion artifacts for unknown but present motion patterns. As part of this research, it was discovered that image blurring due to motion is present extensively in 3D breast CT images, even without it being apparent to breast radiologists. The developed method can remove this motion artifact, improving the image quality in a substantial portion of cases.
Finally, a reconstruction method to quantify and reconstruct the iodine concentration throughout the imaged breast was implemented and initially tested. This method seems to recover the iodine content accurately in basic simulations, and is now being extended to be able to reconstruct 4D breast CT iodine images.
Various stages of 4D breast CT image acquisition have been developed, all of them representing progress beyond the state of the art, since 4D breast CT is a new modality, being developed as part of this project. These involve the algorithms to correct for x-ray scatter and motion, and the development of a model of 4D breast CT that will be used to optimize the acquisition parameters for this modality. Also, the development of dynamic phantoms, both with movement of solid components or with fluid flow through the static representation of a breast, are contributions that go beyond the current state of the art.
The rest of the project will involve the use of the breast CT model to find the optimal settings for the acquisition of 4D breast CT images, the development and optimization of 4D reconstruction of iodine images, and the final development of the dynamic fluid-flow phantoms.
Once all stages of the image formation process are optimized, understood, and thoroughly tested, patient imaging for both treatment planning and monitoring will be performed and tested.
We expect that by the end of the project the complete process of 4D breast CT imaging will have been developed, optimized, tested, and optimized, and that the first patient trials will demonstrate its feasibility for the target clinical applications.
The rest of the project will involve the use of the breast CT model to find the optimal settings for the acquisition of 4D breast CT images, the development and optimization of 4D reconstruction of iodine images, and the final development of the dynamic fluid-flow phantoms.
Once all stages of the image formation process are optimized, understood, and thoroughly tested, patient imaging for both treatment planning and monitoring will be performed and tested.
We expect that by the end of the project the complete process of 4D breast CT imaging will have been developed, optimized, tested, and optimized, and that the first patient trials will demonstrate its feasibility for the target clinical applications.