High-Precision Intelligent Cancer Diagnosis Method: 2D Raman Figures Combined with Deep Learning

Abstract

Raman spectroscopy, as a label-free detection technology, has been widely used in tumor diagnosis. However, most tumor diagnosis procedures utilize multivariate statistical analysis methods for classification, which poses a major bottleneck toward achieving high accuracy. Here, we propose a concept called the two-dimensional (2D) Raman figure combined with convolutional neural network (CNN) to improve the accuracy. Twodimensional Raman figures can be obtained from four transformation methods: spectral recurrence plot (SRP), spectral Gramian angular field (SGAF), spectral short-time Fourier transform (SSTFT), and spectral Markov transition field (SMTF). Two-dimensional CNN models all yield more than 95% accuracy, which is higher than the PCA-LDA method and the Raman-spectrum-CNN method, indicating that 2D Raman figure inputs combined with CNN may be one reason for gaining excellent performances. Among 2D-CNN models, the main difference is the conversion, where SRP is based on the structure of wavenumber series with the best performances (98.9% accuracy, 99.5% sensitivity, 98.3% specificity), followed by SGAF on the wavenumber series, SSTFT on wavenumber and intensity information, and SMTF on wavenumber position information. The inclusion of external information in the conversion may be another reason for improvement in the accuracy. The excellent capability shows huge potential for tumor diagnosis via 2D Raman figures and may be applied in other spectroscopy analytical fields.

Practices, Potential, and Perspectives for Detecting Predisease Using Raman Spectroscopy

“…Furthermore, surface-enhanced Raman scattering (SERS), spatially offset Raman spectroscopy (SORS), and incorporating machine learning are complementary technologies that increase the detection sensitivity in liquid biopsies [63][64][65] and deep layers of tissues and organs [58,66]. They also increase accuracy in discriminant analysis [67][68][69][70][71][72].…”

“…The literature involving Raman spectroscopy and human diseases is too numerous to list so representative articles are selectively cited here (Table 1). For the early detection and prediction of human disease, applications range from the discriminant analysis of cancer cells [72,94], tissues [51,53,62,69,71,[95][96][97], and serum sample [67,98] to diagnostic procedures via endoscopy [95]. Cheng et al demonstrated that four leukocyte types (granulocytes, monocytes, B cells, and T cells) from healthy people were characterized as a reference of normal hematopoiesis and were distinguished from each other by generating an orthogonal partial least squares discriminant analysis (OPLS-DA) model for the further analysis of leukemic granulocytes [72].…”

Accurate and rapid molecular subgrouping of high-grade glioma via deep learning-assisted label-free fiber-optic Raman spectroscopy

“…Raman spectroscopy, a label-free optical vibrational spectroscopy technique, has been widely used to quantitatively analyze molecular compositions of biological specimens and diagnosis of various human cancers ( 13–21 ). AI algorithms have been shown to improve the robustness of Raman spectroscopy-based precise cancer diagnosis ( 22 ). Particularly, previous studies have demonstrated the capability of Raman spectroscopy in intraoperative brain cancer detection ( 23–27 ).…”

Accurate and Rapid Detection of Peritoneal Metastasis from Gastric Cancer by AI‐Assisted Stimulated Raman Molecular Cytology

“…[36][37][38] AI algorithms improved the robustness of Raman spectroscopy for high-precision cancer diagnosis. [39] However, due to the weak spontaneous Raman signals, Raman spectroscopy-based cytology took up to 8 h for a complete analysis without spatial information. With remarkably boosted Raman signals, simulated Raman scattering (SRS) microscopy is a desirable method of label-free and high speed molecular imaging at the single cell level.…”