It is well know that proton therapy offers a major advantage compared to the more common photon modalities, because of the finite range of protons in tissue allowing less total dose to the patient and organs at risk (OARs). Given the ability to target tumor volumes, the geometric precision of delivery of proton beam therapy is particularly crucial. Incorrect patient positioning, organ motion, tumor shrinkage, can lead to very high proton doses to OARs and be ineffective to the tumor. 

Our group recently examined the advantages of using proton radiography (PR) showing that PR could provide less dose to the patient, higher soft tissue contrast than X-rays and real time capability for tumor tracking. Although PR offer a better CNR compared to the standard X-ray, the tumor is hardly visible from each BEV for either X-ray or proton radiography, especially from lateral projections going through the mediastinum. 

We developed animage-processing technique to identify the tumor region on proton radiographs in lung cancer patients. The method provides enhanced projective images from the beam’s eye view angle, hence allowing tumor-based set-up and potential tumor tracking.

Tumor Tracking on PR

Multi Atlas Based Segmentation (MABS) is a methodology for automatic image segmentation that 

is proving to be able to provides robust and reliable results.

The rationale behind MABS is to use a set of digital atlases (anatomical image plus labels) and image registration technique to “bring” the the contours from the atlases to the query image. Applying a statistical methodology to all the deformed labels the final segmentation will be extracted. Of course the number and the “type” of the atlases that will be used in this process can significantly influence the quality of the final results and also the time required to get them. In fact it would be advantageous to not include in the workflow the atlases that introduce redundant or noisy information. This filtering process is usually called “atlas selection”.

In our work we are implementing a MABS algorithm into the Plastimatch project 

(www.plastimatch.org). In this software have been introduced many interesting and unique features in order to solve scientific and practical issues related to this type of algorithms. Complete and smart workflow that allows to use data directly from a clinical enviroment, atlas selection based on different strategies, multiple ways for final contour extracting (gaussian and/or staple), fast and accurate image registration process are the main advantages of our implementation.

Several tests are running on head and neck CT dataset (radiotherapy purpose) and brain MRI dataset (neuroscience application) to validate the proposed metodology and its innovative features.

MABS brain

White matter fibers are anatomical structures made by a bundle of thousand of axons.

They represent the “cables” that connect the different areas of the brain and so an accurate identification of these important structures is very important for diagnosis, treatment and to improve our knowledge about brain.

In the last years several algorithms have been proposed in order to segment white matter fibers but a lot of open points remain unsolved.

We are developing a new methodology that aim to improve the current state of the art.

Using our algorithm is it possible to segment those bundles in a fast and accurate manner. Our segmentation is fully automatic, tridimensional, defined on T1 space and the process is extremely fast (a few minutes) if compared to other approaches.

Radiation therapy is one of the two steps for common treatment of head and neck Hodgkin/non­ Hodgkin lymphoma. Unfortunately, in this scenario, clear and accurate localization of tissues to irradiate by radiation is not always easy and obvious, especially after chemotherapy treatment.

This makes the contouring process very challenging and prone to mistakes, even for expert physicians. The purpose of our work is to support the radiotherapists during the definition of clinical target volume for radiotherapy of this kind of cancer.

We propose to localize the gross target volume onto the simulation CT using pre chemotherapy [18F]FDG­PET/CT information and deformable image registration technique. The primitive disease extent brought on the simulation CT can be so used as starting point to easier manually define the whole volume to treat.

As a result of the our tests the methodology proved to perform well, being able to identify the initial cancerous site also in very challenge clinical situations, as enormous difference in patient's position between scans.

Motion of soft tissues is often quantified through wide used ultrasound images, this modality is non-invasive, non-ionizing and have good time resolution. Ultrasound images show a characteristic type of noise called "speckles", due to random interference patterns and natural acoustic reflections. This kind of noise show a certain type of stability through consecutive frames. Identification and tracking of groups of stable speckle allow to quantify displacement of soft tissues.This approach is widely used in ecocardiography in order to estimate atrial or ventricular deformations. We applied this kind of analysis to estimate intima media/adventitia longitudinal displacement in healthy and in subject with stent implanted.

Speckle Tracking

Angiography is wide-used for assessment of peripheal vascular diseases, such as occlusions and neo-angiogenesis. We are looking for new biomarkers capable to give a quantitative information about states of this kind of disorders. Fractal dimension is often used in oncology in order to quantify the capability of a tumor to grow. We're trying to correlate fractal dimensions to degree of growth of new capillaries in response of an occlusion. This challenge include the capability to stitch xray cineangiography in order to build a wide-field angiography. Another interesting task is the accurate segmentation of capillaries from low contrast reconstructed xray angiography.