Deep learning-assisted terahertz intelligent detection and identification of cancer tissue.

Cancer, as one of the most notorious health diseases, represents the main reason attributed to millions of worldwide deaths each year. Timely detection and accurate diagnosis are thus vital to cancer prevention and timely therapy. Traditional cancer prescreening is not only cumbersome but also heavily reliant on sufficient expert knowledge, which inevitably increases the complexity of cancer diagnosis and limits early cancer diagnosis. To overcome this problem, recent terahertz (THz) technology, as an unconventional bio-friendly detection approach, has emerged with great potential in human disease diagnosis due to its non-ionic and high-resolution features. By combining the THz detection technique and artificial intelligence technique, here we propose a dense and efficient channel attention network (DECANet) framework-based THz diagnosis system for cancer prescreening. The cancer identification and diagnosis process are transformed into one end-to-end classification process of THz signals reflected from cancer tissue. The biosamples of breast and skin cancer tissue are characterized to validate the effectiveness and applicability of the proposed approach. Our quantified results indicate that our proposed THz diagnosis framework has promising feature extraction capability for abnormal cancerous tissue and provides an effective complement tool to assist the healthcare cancer diagnosis.