Compressive Sensing in Lung Cancer Images for Telemedicine Application

Irawati, I. D.; Hadiyoso, S.; Fahmi, A.

Telemedicine technology as a solution to prevent the spread of Covid-19. Tele-radiology for lung cancer images requires a large bandwidth when the image is transmitted, whereas the available bandwidth is limited. CT-scan lung cancer image has a very large capacity so that it requires a large storage space, while the storage capacity is very limited. On the sender side, the application of compressive sensing as an alternative solution to obtain data compression with a high compression ratio but requires high accuracy on the receiver. In addition to make it easier for medical staff and doctor for diagnosing the type of lung cancer, the recipient requires a lung cancer image classification, which consists of 3 types of cancer, including: adeno carcinoma (ACA), squamous cell carcinoma (SCC), and benign lung cancer (N). This paper proposes a combination method consisting of a Compressive Sensing (CS) algorithm, feature extraction, and KNN classification that can work effectively and efficiently in telemedicine applications. The results showed that CS worked effectively for compression with large compression ratios without having an influence on the accuracy results. The sparse technique FFT provides the highest accuracy compared to IFFT, DWT and without sparsing. The classification using KNN shows that the N image has uniquely extracted characteristics and give accuracy up to 100%, whereas the image of ACA and SCC provide accuracy by 70%. © 2021 ACM.