Humans in the Loop played a crucial role in the annotation process, which was key to the project’s success. The project involved collecting thermal scans from five different angles of patients’ legs, along with extensive medical data such as demographic information, clinical profiles, and specific medical conditions. This data helped in the identification of peripheral arterial disease (PAD) and other vascular issues.
Humans were essential in annotating over 2,000 graphic images, focusing on segmenting arterial zones in the legs and head. They provided segmentations for 11 arterial zones in the legs, resulting in over 7,000 annotations. This contributed to an algorithm that achieved an 86% segmentation accuracy. Similarly, for the head, humans annotated four arterial zones, leading to 2,000 annotations with a 99% segmentation precision.
These annotations were vital for training machine learning algorithms to accurately classify whether a patient had PAD or other blockages. The algorithms then generated visualizations of the segmented arterial zones, helping specialists determine the patient’s condition and decide on the appropriate next steps in their treatment. This collaboration between human annotators and machine learning significantly enhanced the accuracy and reliability of the diagnostic process.