Table 1.
Summary of the state-of-the-art studies on the use of HSI for cancer analysis.
| Reference | Year | Type of Cancer | Type of Sample | Spectral Range (nm) | Image Size (pixels) | # Bands | Light Source | Acquisition Mode | Algorithms ¥ | Goal | Subject * |
|---|---|---|---|---|---|---|---|---|---|---|---|
| [123] | 2007 | Breast | in-vivo | 450–700 | 1024 × 1528 | 34 | InGaN LEDs | LCTF | Custom Algorithm | Classification | A |
| [142] | 2011 | Oral | in-vivo | 450–650 | 350 × 350 | 48 | Halogen | Snapshot | - | - | H |
| [143] | 2011 | Oral | in-vivo | 400–700 | - | 40 | Halogen | Snapshot | PCA, LDA | Dimensional reduction, Classification |
H |
| [116] | 2011 | Gastric | ex-vivo | 1000–2500 | - | 239 | Halogen | Pushbroom | SVM, Integral Method, NDCI | Classification, Margin delineation |
H |
| [184] | 2012 | Prostate | in-vivo | 450–950 | 1392 × 1040 | 251 | Xenon | LCTF | LS-SVM | Classification | A |
| [146] | 2012 | Tongue | in-vivo | 600–1000 | 1392 × 1040 | 81 | Halogen | AOTF | SR, SVM, RVM | Classification | H |
| [185] | 2012 | Prostate | in-vivo | 500–950 | 1392 × 1040 | 251 | Xenon | LCTF | LS-SVM | Classification | A |
| [117] | 2013 | Gastric | ex-vivo | 400–800 | 640 × 480 | 72 | Halogen | - | Cutoff point | Optimal wavelength selection, Classification |
H |
| [125] | 2013 | Breast | ex-vivo | 380–720 | - | 101 | Xenon | - | Polynomial SVM | Automatic ROI detection based on contrast and texture information | H |
| [126] | 2013 | Breast | ex-vivo | 380–720 | - | 101 | Xenon | - | Fourier coefficient selection features, mRMR, RBF SVM |
Feature extraction, Dimensional reduction, Classification |
H |
| [124] | 2014 | Breast | in-vivo | 500–600 | 1392 × 1040 | 26 | Halogen | LCTF | Gabor Filter, Expectation Maximization |
Microvessel sO2 segmentation & classification | A |
| [132] [133] |
2014 | H&N | in-vivo | 450–950 | 1392 × 1040 | 251 | Xenon | LCTF | Tensor Decomposition, PCA, KNN |
Feature extraction, Classification |
A |
| [134] | 2014 | H&N | in-vivo | 450–950 | 1392 × 1040 | 251 | Xenon | LCTF | PCA, FFD | Surgical margin delineation and in-vivo/in-vitro registration | A |
| [136] | 2015 | H&N | in-vivo | 450–950 | 1392 × 1040 | 226 | Xenon | LCTF | mRMR, KNN | Glare removal, Feature extraction, Automatic classification |
A |
| [135] | 2015 | H&N | in-vivo | 450–950 | 1392 × 1040 | 226 | Xenon | LCTF | mRMR, RBF SVM, Chan-Vase active contour method |
Glare removal, Feature extraction, Automatic classification, Active contour refinement |
A |
| [118] | 2015 | Gastric | ex-vivo | 400–800 | 480 × 640 | 81 | Halogen | - | Mahalanobis distance, Cutoff point |
Optimal wavelength selection, Classification |
H |
| [145] | 2016 | Oral | in-vivo | 390–680 | - | 30 | - | - | RF | Classification | H |
| [78] | 2016 | Oral | in-vivo | 390–680 | 1388 × 1040 | 30 | Xenon | - | Customized | Image filtering (honeycomb pattern removal) | H |
| [120] | 2016 | Colon | in-vivo | 405–665 | 585 × 752 | 27 | Xenon | Filter Wheel | Recursive divergence, SVM | Wavelength selection, Classification |
H |
| [137] | 2016 | H&N | in-vivo | 450–950 | 1392 × 1040 | 251 | Xenon | LCTF | SVM, MSF | Classification & segmentation | A |
| [144] | 2016 | Oral | in-vivo | 390–680 | 1388 × 1040 | 30 | Xenon | - | NCC, MNF, RF |
Image registration and denoising, Glare detection, Classification |
H |
| [140] | 2017 | H&N | ex-vivo | 450–950 | 1392 × 1040 | 91 | Xenon | LCTF | CNN, SVM, KNN, LR, DTC, LDA |
Classification | H |
| [138] | 2017 | H&N | ex-vivo | 450–50 | 1392 × 1040 | 91 | Xenon | LCTF | Ensemble LDA | Classification | H |
| [139] | 2017 | H&N | ex-vivo | 450–950 | 1392 × 1040 | 91 | Xenon | LCTF | LDA, QDA, Ensemble LDA, Linear SVM, RBF SVM, RF |
Classification | H |
| [119] | 2019 | Colon | ex-vivo | 400–1000 900–1700 |
1 × 1312 1 × 320 |
- | Halogen | Pushbroom | Quadratic SVM | Classification | H |
| [186] | 2019 | H&N | ex-vivo | 450–950 | 1392 × 1040 | 91 | Xenon | LCTF | Inception CNN | Binary and Multiclass Classification | H |
| [173] | 2016 | Brain | in-vivo | 400–1000 900–1700 |
1 × 1004 1 × 320 |
826 172 |
Halogen | Pushbroom | SVM, RF, ANN |
Classification | H |
| [169] | 2016 | Brain | in-vivo | 400–1000 | 1 × 1004 | 826 | Halogen | Pushbroom | RF | Pre-Processing and Classification | H |
| [170] | 2017 | Brain | in-vivo | 400–1000 900–1700 |
1 × 1004 1 × 320 |
826 172 |
Halogen | Pushbroom | tSNE, FR-tSNE STF, DCT-STF |
Dimensional Reduction and Classification | H |
| [164] [171] |
2018 | Brain | in-vivo | 400–1000 | 1 × 1004 | 826 | Halogen | Pushbroom | SVM, FR-tSNE/PCA, KNN Filter, K-Means, MV |
Classification | H |
| [165] [166] |
2019 | Brain | in-vivo | 400–1000 | 1 × 1004 | 826 | Halogen | Pushbroom | CNN, DNN, SVM, KNN Filter, K-Means |
Binary and Multiclass Classification | H |
* Subject: (H) Human; (A) Animal. ¥ Algorithms: (PCA) Principal Component Analysis; (LDA) Linear Discriminant Analysis; (SVM) Support Vector Machine; Normalized Cancer Index (NDCI); (LS-SVM) Least-Squares Support Vector Machine; (SR) Sparse Representation; (RVM) Relevance Vector Machine; (mRMR) maximal Relevance and Minimal Redundancy; (RBF) Radial Basis Function; (RF) Random Forest; (MSF) Minimum-Spanning Forest; (NCC) Normalized Cross-Correlation; (MNF) Minimum Noise Fraction; (CNN) Convolutional Neural Network; (KNN) K-Nearest Neighbor; (LR) Linear Regression; (DTC) Decision Tree Classification; (QDA) Quadratic Discriminant Analysis; (ANN) Artificial Neural Network; (tSNE) t-Distributed Stochastic Neighbor Embedding; (FR-tSNE) Fixed Reference t-Distributed Stochastic Neighbor Embedding; (STF) Semantic Texton Forests; (DCT-STF) Discrete Cosine Transform based Semantic Texton Forest; (MV) Majority Voting; (DNN) Deep Neural Network.