Patient and Treatment Monitoring #
Stefano Trebeschi, Teresa Tareco Bucho, Melda Yeghaian, Kevin Groot Lipman, Laura Jovani Estacio Cerquin, Melissa de Bruin, Iris van der Loo
What is patient monitoring in cancer care? #
Patient monitoring is an umbrella term, used for all diagnostic tests and exams performed by doctors and health care professionals. It ensures that the patient, who is receiving anti-cancer treatment, is doing well, that the treatment is effective, and that their conditions are improving or, at least, stable. Terms commonly used in patient monitoring include follow-up and response assessment. Follow-up refers specifically to the set of regularly scheduled exams that the patient undergoes during or after treatment. Response assessment refers to the method used to define treatment effect on the patient: simply put, to answer the question “is the cancer shrinking?”.
What is cancer progression? #
When the treatment is ineffective, or fails to slow down the disease, we talk about disease progression. There is no universal test to define what progression is.
If the tumor, or parts of it, are taken out, for example by means of biopsy, we can evaluate it under the microscope. If there is an increase in the number and location of cancer cells, we can call that pathological progression.
If a radiological image is available, that is a scan/image of the internal anatomy of the patient; we can measure if the volume of the tumor seen across the body has increased. We call that a radiological progression.
If the symptoms or signs of the disease have increased, observable on the patient by their treating physician, that is called a clinical progression.
There are other specific methods used to measure response to treatment, which also have their own definition of progression, depending on the cancer type, the rules used to assess it, and the diagnostic test: RECIST progression, RANO progression, GLEASON grade progression, progression according to TNM restaging, etc.
Why are there so many definitions of progression? #
Every diagnostic test for progression has its limitations.
For example, pathological progression cannot produce accurate results in advanced cancer patients, where the tumor has spread in multiple locations, and only one/two locations can be tested; or in patients where the tumor is just not reachable by biopsy. On top of that, it is rarely possible to observe the whole tumor under a microscope.
With radiological progression, which can only look at the overall anatomy, it is not possible to identify tumor cells, and we have to rely on the fact that each tumor (or just lump) seen in the scan has the same density of cancer cells. Each “tumor-lump” (a.k.a. metastases) however, can pack different types of cells, with different densities, depending on the environment and circumstances it finds itself in. These differences can also bring each metastases to grow and react differently to the same treatment, within the same patient, making it difficult to define unequivocally what an increase in volume looks like.
On top of that, many of these tests are highly relying on the experience and knowledge of the diagnostician interpreting the test results. This means that often, for many of these tests, two diagnosticians may end up with two different answers for the same patient (e.g. progression vs non progression).
Is this a problem? #
Having different ways to measure progression is not, per se, a problem, as long as these are integrated: used together. This however rarely happens, with most clinical trials and clinical practice relying on one criteria.
The variability and lack of reproducibility of most response evaluation criteria however is a problem. This introduces statistical noise in trial data, depending on the method chosen, resulting in more patients required per trial. On top of that, misclassifying or inability to recognize progression in a timely manner can result in continuing ineffective treatment with the unnecessary side-effects.
Where do we go from here? #
We need new tools to measure response of patients, and objectively define progression.
Ideally, we would be able to track all cancer cells across the whole body, and their effect on the health of the patient. Practically speaking, we will have to develop proxy tools: tools that can estimate the tumor cell count and patient health.
As any other measuring device used in science and engineering, these tools have to be reproducible and repeatable, and not subject to a personal interpretation of the operator.
This research line is tackling exactly this: measuring the reproducibility and objectivity of existing tools for response evaluation, development of new tools to improve the limitations of current ones, and optimizing their integrative usage.
Published work #
Stefano Trebeschi et al. “Prognostic value of deep learning mediated treatment monitoring in lung cancer patients receiving immunotherapy”. In: Frontiers in Oncology, 10.3389/fonc.2021.609054 (2021). link;
Stefano Trebeschi et al. “Development of a prognostic AI-monitor for metastatic urothelial cancer patients receiving immunotherapy”. In: Frontiers in Oncology, 10.3389/fonc.2021.637804 (2021). link;
Ingmar Paul Loohuis “Exploring the prognostic value of image-to-image registration for immunotherapy patient monitoring”. MSc thesis, Technical Medicine, University of Twente (2022). link;
Iris van der Loo “Prognostication from Longitudinal Multisequence Brain MRI using Artificial Intelligence”. MSc thesis, Technical Medicine, University of Twente (2022). link;
Kevin Groot Lipman et al. “Artificial intelligence-based diagnosis of asbestosis: analysis of a database with applicants for asbestosis state aid”. Eur Radiol. 2022 Dec 26. doi: 10.1007/s00330-022-09304-2. PMID: 36567379. link;
Christof Schulz et al. “Design and Evaluation of a Prototype for a Platform for AI algorithms in Medical Imaging: A Human-Centered Approach”. MSc thesis, Human Factors and Engineering Psychology, University of Twente (2023) link not yet available.