2-Phase Framework of Suspicious Region Identification in Medical Thermal Images

Thermography-based technology is an ecofriendly approach of screening abnormality in human bodies. Thermal cameras complemented with robust and automated Computational Intelligence Aided Diagnostic Systems can be deployed at public gathering points to warn about thriving diseases in the human body due to inflammation, thereby strengthening human health systems. In this paper, we have proposed a two-phase framework, wherein phase 1 (Classification Phase) characterizes and classifies abnormal thermal patterns using deep learning techniques and phase 2 (Segmentation Phase) segments the affected region in human body using Fuzzy clustering approach. Also, this paper reviewed and compared the state-of-the-art methods for classifying and segmenting abnormality in medical thermal images. We showed the effectiveness of our framework. The experimental performance shows that the proposed framework is pragmatic and result-oriented. We obtained an accuracy of 91.33%, 89.5% and 93.67% on three datasets – DB-FOOR-IR, DB-THY-IR and DB-DMR-IR, respectively for the classification phase. In phase 2, we obtained Uniformity indices as 0.905, 0.811, and 0.861 for DB-DMR-IR, DB-FOOT-IR, and DB-THY-IR datasets, respectively.