Development and Validation of a Deep Learning-Based Infrared Thermography Model for Detection and Differentiation of Thyroid Nodules

Patients referred for thyroid nodule (TN) assessment will be invited to participate in the study. Patients will undergo screening using infrared thermography (IT) and ultrasound (US) with needle insertion. Neck IT thermograms will be captured using the FLIR TS865 for all patients’ frontal and lateral norms. With the IT camera software, the regions of interest (ROI) will be defined at the thyroid glands. The reference examination will label the IT images according to the following classifications: 1) control, 2) TN, and 3) TN cancer group. The IT images will undergo enhancement and segmentation using Active Contour without Edge (ACWE). The segmented images will build a convolutional neural network (CNN) that utilizes features extracted from the acquired IT images for detection and differentiation. The CNN will then be tested and validated for its use as the first screening tool for thyroid nodules.

Faculty Supervisor:

Daniela de Melo

Student:

Partner:

Universidade Estadual de Maringá

Discipline:

Life Sciences

Sector:

Artificial Intelligence; Health and Related Sciences & Technology; Technology

University:

University of Saskatchewan

Program: