Artificial intelligence based real-time microcirculation analysis system for laparoscopic colorectal surgery

Figure 2 Block diagram of indocyanine green curve pattern classification model. Overview of processing algorithm including training process, indocyanine green (ICG) curve pattern classification process and risk evaluation process. Training data set is acquired through pre-processing from ICG videos of 50 patients and it consists of 10000 ICG curves. To obtain a learned self-organizing map network, the input data set is classified into 25 patterns, and risk labeling is performed on each classified pattern. When a new ICG curve comes in, the learned network classifies it into the most similar pattern and evaluates risk. ICG: Indocyanine green; SOM: Self-organizing map.