Table 1.

Past studies related to cancer detection using artificial intelligence (AI) techniques.

Author	AI Techniques	Classification Types	Parameters	Performance
Mohanty et al., 2011 [11]	Association Rule	Binary (benign vs. malignant)	GLCM and GLRLM features	The association rule method was used for image classification. Accuracies of 94.9% and 92.3% were achieved using all and significant features, respectively.
Neeta et al., 2015 [12]	SVM and K-NN	Binary (benign vs. malignant)	GLDM and Gabor features	Detection of PCa using CAD algorithm and the best result of 95.83% was achieved using an SVM classifier.
Filipczuk et al., 2012, [13]	K-NN	Binary (benign vs. malignant)	GLCM and GLRLM features	Breast cancer diagnosis was performed by classifying the texture features based on GLCM and GLRLM extracted from the segmented nuclei. The best result of 90% was obtained by combining the optimal features of GLRLM.
Radhakrishnan et al., 2012 [14]	SVM	Binary (benign vs. malignant)	Histogram, GLCM, and GLRLM features	TRUS medical images were used for prostate cancer classification. The DBSCAN clustering method was used for extracting the prostate region. The best accuracy of 91.7% was achieved by combining the three feature extraction methods.
Sinecen et al., 2007 [15]	MLP1, MLP2 RBF, and LVQ	Binary (benign vs. malignant)	Image texture based on Gauss-Markov random field, Fourier transform, stationary wavelets	Prostate tissue images of 80 benign and 80 malignant cell nuclei were evaluated. The best accuracy of 86.88% was achieved using MLP2.
Bhattacharjee et al., 2019 [16]	MLP	Binary (benign vs. malignant)	Color moment and GLCM features	Wavelet-based GLCM and color moment descriptor were extracted from prostate tissue images of benign and malignant classes. The model achieved an accuracy of 95%.
Song et al., 2018 [17]	DCNN	Binary (cancer vs. non-cancer)	MRI scans	PCa and noncancerous tissues were distinguished using DCNN. An AUC of 0.944, a sensitivity of 87.0%, a specificity of 90.6 PPV of 87.0%, and an NPV of 90.6% were achieved using DCNN.
Bhattacharjee et al., 2019 [18]	SVM	Binary (benign vs. malignant) and Multiclass (benign vs. grade 3 vs. grade 4 vs. grade 5)	Morphological features	Morphological feature classification was performed for discriminating benign from malignant tumor, grade 3 from grade 4, 5 tumors, and grade 4 from grade 5 tumor. The best rest was obtained from binary classification.
Zhao et al., 2015 [19]	ANN	Binary (PCa vs. non-Pca)	GLCM, Gray-level histogram, and general features	T2-weighted prostate MRI scans were used to extract 12 different types of features. Feature classification was performed using an artificial neural network. For PZ and CG, the accuracies achieved using a CAD system were 80.3% and 84.0%, respectively.
Roy et al., 2019 [20]	CNN	Binary (nonmalignant vs. malignant) and Multiclass (normal vs. benign vs. in situ vs. invasive carcinoma)	Histology images	The patch-based classifier using CNN was developed for the automated classification of histopathology images. In classifying the images of the cancer histology test dataset, the proposed technique achieved promising accuracies for both binary and multiclass classification.
Chakraborty et al., 2020 [21]	DCRCNN	Binary (cancer vs. noncancer)	Histopathologic scans	A dual-channel residual convolution neural network was used to classify the tissue images of the lymph node section. The model was trained with 220,025 images and achieved an overall accuracy of 96.47%.