Automated tissue classification system for cancer detection and image-guided procedures.
- Allows fast and precise diagnoses of tissue abnormalities including tumors.
- Collects both spatial and spectral data of tissue.
- Provides multispectral imaging to extract and visualize additional information the human eye cannot detect.
- Captures numerical data for patient follow-up and tracking of small changes in patients.
Hyperspectral imaging extends human vision to near-infrared and infrared wavelength regions and can provide a powerful tool for non-invasive tissue analysis. It is already used to measure oxygen saturation in patients with peripheral vascular disease, to detect ischemic regions of the intestine during surgery, to predict and follow healing in foot ulcers of diabetic patients, and to diagnose hemorrhagic shock. Hyperspectral imaging can also discriminate between non-cancerous and cancerous tissue. The current gold standard for cancer diagnoses is biopsy followed by analysis of the tissue via histopathology in the lab, a process that can take several days. Hyperspectral imaging has never been used for cancer detection in pathological slides.
With the use of hyperspectral imaging, Emory researchers have developed an optical system that can differentiate tumors from normal tissue for the detection of cancer as well as for image-guided procedures such as surgery and interventions. The system includes a broad-band light source selected to illuminate the tissue glass slide and a hyperspectral camera to capture all wavelength bands from 450 to 950 nm. The system can capture both the spatial and spectral data of tissue and has been trained to classify tissue on histologic slides based on predetermined pathology using lung tissue and lymph nodes from mouse models that were transplanted with highly metastatic human head and neck cancer cells. After training, this technology can detect metastatic cancer with a specificity of 97.7% and sensitivity of 92.6% in lung tissue and a specificity of 98.3% and sensitivity of 96.2% in lymph nodes. This system could help pathologists evaluate more histologic slides accurately, reliably, and in less time.
Imaging software has been tested on resulting histologic slides from mouse models for head and neck cancer with very high specificity and sensitivity.
Publication: Akbari et al., Proc. SPIE (Vol 8317) Medical Imaging 2012