Deep Learning in Cancer Diagnosis and Prognosis Prediction: A Minireview on Challenges, Recent Trends, and Future Directions

Table 5

Summary of the studies for the classification of other types of cancer.

Publication	Type(s) of cancer	Type of data	Methods	Performance
[125]	Colorectal cancer	Custom dataset	VGG19, AlexNet, SqueezeNet version 1.1, GoogLeNet, ResNet-50	98.7% accuracy using VGG19 model
[126]	Colon cancer	Custom dataset	CNN model	, ,
[127]	Gastric cancer	Custom datasets	DL models	Accuracy of 0.822 in the international validation cohort
[128]	Bladder cancer	TCGA+custom dataset	DL models
[129]	Lung cancer	TCGA+custom dataset	Weakly supervised DL algorithm	Accuracy of 97.3% on the custom dataset
[130]	Prostate cancer	Custom dataset	DL methods	Accuracy of 92% in cancerous/benign classification
[131]	Skin cancer	Custom dataset	DL algorithms	Positive predictive value of 59.9%
[132]	Skin cancer	HAM10000, Kaohsiung Chang Gung Memorial Hospital	Lightweight DL algorithm	(HAM10000, multiclass classification)
[133]	Skin cancer	Custom dataset	Interpretable DL methods	Accuracies between 93.6% and 97.9%
[134]	Skin cancer	Custom dataset	CNN model	Accuracy of 82.95% for multiclass classification
[135]	Liver cancer	TCGA dataset	DL model	High accuracy (abnormal/normal classification)
[136]	Head and neck cancer	Custom dataset	Inception version 3	Mean AUC was 0.936 based on the testing set
[137]	Pancreatic cancer	Three custom datasets	CNN architectures	Accuracy of 0.986 for test set 2
[138]	Multiple	Custom datasets	ResNet-18, ResNet-34, ResNet-50	Accuracy of 94.90% for ResNet-50 architecture
[139]	Breast cancer	MITOS12, 2016 Tumor Proliferation Assessment Challenge	CNN architecture	-measure of 0.79
[140]	Multiple	Custom dataset	Multiple instance learning-based DL system
[141]	Blood and bone marrow cancer	Leukemia microarray gene data, Gene Expression Omnibus repository	Single-layer neural network, 3-layered deep network	96.67% for 3 layered model
[142]	Multiple	BioGPS data portal, TCIA, GDC dataset	Regression-based partitioned DL algorithm
[143]	Multiple	TCGA, 1000 Genomes Project	DL algorithms
[144]	Multiple	Kvasir dataset, Gastrolab	DL-based classification network
[145]	Multiple	Custom datasets	CNN model
[146]	Multiple	Custom dataset	3D CNN model
[147]	Multiple	Custom dataset	GAN-based model	90–99% accuracies