HOLY-2020

Cancer cells need a lot of sugar (also: glucose) for their rapid replication and the growth of a tumour. It is possible to visualize the sugar consumption by an imaging technique called Positron Emission Tomography (PET). This is done by providing the patient with minuscule amounts of radioactively labelled molecules of sugar; their path inside the body can then be tracked from the outside within a few minutes.

PET is well-established in the diagnosis and treatment monitoring of Hodgkin‘s lymphoma (HL).  HL is a type of tumour that can be characterized by its sugar consumption. Thus, PET enables early monitoring of treatment response, that is shows if the lymphoma shrinks or grows, reflecting whether the therapy is working or not working. So far, only very basic information of the acquired PET images is used. Since the biology of HL is linked to its metabolism, we aim to analyze PET images in more detail by using artificial intelligence algorithms. By doing so, we seek to identify those patients specifically who suffer from aggressive HL and who are in need of more intense treatment. The same algorithms can hopefully be used also to identify patients with a favourable prognosis and who require less intense treatments with fewer side effects. 

All in all, our anticipated combination of PET imaging with AI postprocessing holds the potential for personalized treatment planning with immediate and long-term benefits to patients.

We aim to validate established algorithms using retrospective FDG PET/CT image data, clinical data and machine learning (ML)/deep learning (DL) for treatment stratification in early stage Hodgkin lymphoma (HL). The results of this project will be the basis for prospective studies, which will pave the way for customized therapies with fewer side effects and a better quality of life for the patients. Despite of FDG PET/CT being already an integral part of initial staging and treatment monitoring in lymphoma patients, standard prognostic scores fail to reliably stratify patients. In particular, the Deauville score fails to identify patients who do not need intensified treatment with detrimental late side effects in a predominantly adolescent patient population. It is agreed that not only malignant lesions, but the entire organism must be investigated in a systemic way in order to correctly characterize a disease. ML and radiomics models have proven potential in prognostic stratification in lymphoma. However, they mostly analyze the main site of lymphoma-involvement only, without accounting for additional factors such as features associated with immune response to the disease or disease-induced imaging features throughout the body. Given the complex and highly variable distribution of lesions throughout the body of HL patients, and the variety of imaging-based, biological and clinical prognostic factors on hand, this disease is a model of choice to develop a systems medicine approach. This consortium seeks to effectively combine already proven prognostic factors including metabolic features from state-of-the-art FDG PET/CT imaging with clinical data such as genetic subtype, parameters included in the International Prognostic Score, and patient characteristics in order to validate a clinical decision support system that is capable of stratifying early stage HL with high accuracy.

HOLY-2020

Cancer cells need a lot of sugar (also: glucose) for their rapid replication and the growth of a tumour. It is possible to visualize the sugar consumption by an imaging technique called Positron Emission Tomography (PET). This is done by providing the patient with minuscule amounts of radioactively labelled molecules of sugar; their path inside the body can then be tracked from the outside within a few minutes.

PET is well-established in the diagnosis and treatment monitoring of Hodgkin‘s lymphoma (HL).  HL is a type of tumour that can be characterized by its sugar consumption. Thus, PET enables early monitoring of treatment response, that is shows if the lymphoma shrinks or grows, reflecting whether the therapy is working or not working. So far, only very basic information of the acquired PET images is used. Since the biology of HL is linked to its metabolism, we aim to analyze PET images in more detail by using artificial intelligence algorithms. By doing so, we seek to identify those patients specifically who suffer from aggressive HL and who are in need of more intense treatment. The same algorithms can hopefully be used also to identify patients with a favourable prognosis and who require less intense treatments with fewer side effects. 

All in all, our anticipated combination of PET imaging with AI postprocessing holds the potential for personalized treatment planning with immediate and long-term benefits to patients.

We aim to validate established algorithms using retrospective FDG PET/CT image data, clinical data and machine learning (ML)/deep learning (DL) for treatment stratification in early stage Hodgkin lymphoma (HL). The results of this project will be the basis for prospective studies, which will pave the way for customized therapies with fewer side effects and a better quality of life for the patients. Despite of FDG PET/CT being already an integral part of initial staging and treatment monitoring in lymphoma patients, standard prognostic scores fail to reliably stratify patients. In particular, the Deauville score fails to identify patients who do not need intensified treatment with detrimental late side effects in a predominantly adolescent patient population. It is agreed that not only malignant lesions, but the entire organism must be investigated in a systemic way in order to correctly characterize a disease. ML and radiomics models have proven potential in prognostic stratification in lymphoma. However, they mostly analyze the main site of lymphoma-involvement only, without accounting for additional factors such as features associated with immune response to the disease or disease-induced imaging features throughout the body. Given the complex and highly variable distribution of lesions throughout the body of HL patients, and the variety of imaging-based, biological and clinical prognostic factors on hand, this disease is a model of choice to develop a systems medicine approach. This consortium seeks to effectively combine already proven prognostic factors including metabolic features from state-of-the-art FDG PET/CT imaging with clinical data such as genetic subtype, parameters included in the International Prognostic Score, and patient characteristics in order to validate a clinical decision support system that is capable of stratifying early stage HL with high accuracy.