Tumor tissue classification based on micro-hyperspectral technology and deep learning

Abstract

In order to explore the application of hyperspectral technology in the pathological diagnosis of tumor tissue, we used microscopic hyperspectral imaging technology to establish a hyperspectral database of 30 patients with gastric cancer. Based on the difference in spectralspatial features between gastric cancer tissue and normal tissue in the wavelength of 410-910 nm, we propose a deep-learning model-based analysis method for gastric cancer tissue. The microscopic hyperspectral feature and individual difference of gastric tissue, spatial-spectral joint feature and medical contact are studied. The experimental results show that the classification accuracy of proposed model for cancerous and normal gastric tissue is 97.57%, the sensitivity and specificity of gastric cancer tissue are 97.19% and 97.96% respectively. Compared with the shallow learning method, CNN can fully extract the deep spectral-spatial features of tumor tissue. The combination of deep learning model and micro-spectral analysis provides new ideas for the research of medical pathology.

Semi-automatic decision-making process in histopathological specimens from Barrett’s carcinoma patients using hyperspectral imaging (HSI)

“…The sensitivity values of the separate classes EAC and background were up to 25% higher than for stroma and squamous epithelium. In gastric cancer specimens a classification was done based on micro-hyperspectral technology using a deep-learning modelbased analysis method [7]. The experimental results showed a sensitivity with 96% for cancerous and normal gastric tissue for 30 patients.…”

“…The experimental results showed a sensitivity with 96% for cancerous and normal gastric tissue for 30 patients. In contrast to our work, the whole image was annotated in [7], so more spectral data for each patient were available. With a higher number of spectral data, we assume that a deep learning network could improve the results further.…”

“…This was shown in several in-vivo and ex-vivo studies [2]- [5]. In [6], [7] algorithms to distinguish cancerous and non-cancerous tissue of gastric and pancreatic tumors were applied on stained pathological slides.…”

Surgical spectral imaging

“…Artificial neural networks (ANNs) and, more specifically, deep learning with CNNs have recently seen growth in detection, classification and segmentation problems in a variety of medical imaging modalities due to their efficient architecture and use of local context ( Shen et al., 2017 ). Recent applications in spectral imaging analysis include use of CNNs to differentiate abdominal organs ( Akbari et al., 2008b ), and detect gastric ( Hu et al., 2019 ) and head-and-neck cancers ( Halicek et al., 2017 , 2019 ). Reported accuracy in these studies is high (>95%) and a recent comparative study also suggests that CNNs may out-perform competing supervised classification methods, such as SVMs ( Halicek et al., 2017 ).…”

Deep Learning Analysis of in Vivo Hyperspectral Images for Automated Intraoperative Nerves Detection

“…HSI technology has been previously applied in digestive surgery to quantify intestinal perfusion before anastomosis during several procedures [6][7][8][9], as well as in case of mesenteric ischemia [10,11], or to quantify liver perfusion [12]. A number of previous works focused successfully on the ability of HSI to discriminate between normal and tumoral tissue, in particular prostate cancer [13], colorectal cancer [14,15], gastric cancer [16,17], glioblastoma [18] and head and neck cancers [19][20][21][22]. In the oncological eld, advances in hyperspectral signature classi cation have been remarkable and lead to the successful use of sophisticated deep learning algorithms [16-19, 21, 22].…”