Published on August 25, 2021 – Updated on August 25, 2021

Quantitative assessment of therapeutic response or recurrence diagnosis using PET or MRI biomarkers Lesion detectability improvement on CT images using mathematical models Automatic lesion detectability using deep learning program based on artificial intelligence.

X-ray computed tomography : Image quality (noise) assessment using an anthropomorphic phantom for different iterative algorithm level.

X-ray computed tomography : Image quality (noise) assessment using an anthropomorphic phantom for different iterative algorithm level.

Functional imaging such as positron-emission tomography (PET) or magnetic resonance imaging (MRI) have demonstrated a high potential for assessment of therapeutic response or recurrence diagnosis.

In MRI, we have developed a reliable and reproducible method based on dynamic T1-MRI relaxometry measurements to accurately determine physiological tissue parameters (the vascular permeability (kep), the elimination rate constant (K) and the volume fraction (ve) of the extracellular extravascular space) which reflects the tumour micro-vascularisation.The predictive value for local brain metastasis control of a combined score of physiological tissue parameters measured with the developed dynamic T1 relaxometry MRI method already evaluated in different pathologies (breast, cervical tumor) will be evaluated on patients treated by preoperative stereotactic radiotherapy.

In radiology, new X-ray computed tomography (CT) allow, with the development of iterative reconstruction (IR) algorithms, to maintain image quality while significantly reducing patient radiation exposure and thus meet a major challenge of dose reduction while improving lesion detectability. However, safe optimization of CT protocols that use these IR algorithms requires some understanding of their implementation. In particular, in accordance with Report 54 of the International Commission on Radiation Units and Measurements, image quality in CT should be objectively defined as the ability of an observer (human or mathematical) to extract the diagnostic information that gave rise to the image. In this context, we develop a mathematical tool to define the acquisition and reconstruction parameters to obtain the desired diagnostic information at the lowest dosimetric cost and thus improve the detectability of lesions. At the end the performance of our method will be compared with deep learning-based program and visual reading on own database of 211 patients harboring pulmonary nodules with systematic histology available.

Other GCCA Partners