Project workflows

Workflow at the Institute for cancer Genetics and Informatics

Patients are followed up regularly after treatment. Our main interest is to determine whether the patients experience recurrence or not during the follow-up time. For prostate cancer patients, 10 years or more of follow-up is required to put some confidence in our conclusions regarding clinical outcome. This prognosis classification is necessary to develop and evaluate our methods. We split our patient data sets in a discovery data set and a validation data set. The discovery data set is used to iteratively develop methods using the known recurrence classification of patients. In order to evaluate whether the methods work or not, we eventually apply them to new patients in an independent validation data set, i.e. we classify the patients with the method and evaluate whether the classification corresponds with the true clinical outcome. In a clinical setting, our methods are implemented into clinical decision making to tailor treatment for each patient.

Workflow at the ICGI

Workflow in the DoMore! project

A) Slides from the current subject are scanned (A) and de-identified. The coupling between the subject and the anonymized ID are stored inside the context of a hospital for later use. B) The resulting images and the anonymized id are sent to the Cloud for storage, further processing, and analysis. This analysis involves comparing the subject’s data to a generalization of all available data (classification by machine learning), computing well-defined markers and other parameters, which provides insight into the patient’s current health situation and prognosis. C) All information from (B) is available through the electronic health record system (DIPS), which is retrievable from the API using the anonymized id. Clinicians may then review these objective parameters in order to decide correct treatment or follow-up plan. D) The images are rendered in a map-like viewer, with zoom and annotations.

Workflow in the DoMore! project

© DoMore! | Editor-in-chief: Håvard E. Greger Danielsen | Webmaster: Marian Seiergren