gms | German Medical Science

Objectives: Correct scientific image analysis is crucial for counting, grading, scoring or staging but is often hampered by subjectivity and bias. Therefore, image analysis should be performed by multiple blinded observers and in a randomized and anonymized manner. In addition, observers from other research groups should be included using the same standardized tools. In a daily practice however, image analysis is often done by sharing large file folders via e-mail or USB-devices accompanied by an additional spreadsheet for evaluation and later statistical analysis. In a clinical setting, observers are often provided with a list of patients and extract images from the hospital's local PACS, in which patient data are potentially visible. To address all of the abovementioned challenges, we developed Tyche, a web-based tool for an unbiased simultaneous analysis and evaluation of online accessible image collections by multiple observers.

Methods: First, a Tyche session is initiated by a project master who defines a project name, uploads image collections, creates a set of questions with predefined answers and invites colleagues with a project specific URL via an encrypted connection. For the independent observers, Tyche displays images randomly making it impossible to draw any conclusion from the filename. Tyche is equipped with different measuring tools (distances, angles, etc.) and observers can be grouped reflecting their different levels of experience.

Results and Conclusion: 24 histological images of murine knee joint after surgical induction of experimental osteoarthritis (OA) were uploaded. The cartilage degeneration was evaluated at 2, 4, 8, and 12 weeks after OA induction using the OARSI score. Additionally a synovitis score was applied. 4 independent observers with different levels of experience used their personal laptops and finished the analysis in less than an hour. All data was immediately visible for pre-evaluation online and no errors were reported. Results obtained with Tyche confirmed a significant increase in the OARSI score and steady synovitis score after OA induction as reported in the literature.

In order to minimize subjectivity and bias while trying to increase efficiency, we developed Tyche and validated it with histological images after OA induction. In comparison to common procedures, the use of Tyche had several advantages: observers at different worldwide locations can assess images immediately and simultaneously. The sharing of huge data folders is possible and the random and anonymous display of images selected by a project master supports an unbiased analysis. Results from different observers are displayed in real time without any delay. The project master can monitor results online and trace every data point back to the specific observer. We believe that our tool is applicable for almost any other research question that includes measurements, counting and scoring on digital images.