Table 6.

Federated machine learning implementations in cancer prediction.

Ref	Disease	Data Used	Performance
[130]	Brain tumor	Brain MRI Segmentation Kaggle dataset [131]	FL results outperform the baseline but classical ML models competed with their results
[132]	Brain tumor	BraTS dataset [133]	Dice = 0.86 for both FL and ML scenarios
[134]	Brain tumor	BraTS dataset [133]	FL performance is similar to ML models
[135]	Brain tumor	Private data	Dice=0.86 for both FL and ML scenarios
[136]	Skin cancer	ISIC 2018 dataset [137]	Accuracy = 91% for both FL and ML scenarios
[138]	Skin cancer	ISIC 2019 Dermoscopy dataset [137]	Accuracy: 89% which outperformed previous implementations
[139]	Breast cancer	Private data from 7 different institutions	FL perform 6.3% on average better than classical ML
[140]	Breast cancer	Obtained from Netherlands Cancer Registry (NCR)	Not available
[141]	Prostate cancer	Private data	FL model exhibited superior performance and generalizability to the ML models
[142]	Lung cancer	Private data from 8 institutes across 5 countries	Not available
[143]	Pancreatic cancer	Data from hospitals in Japan and Taiwan	FL models have higher generalizability than ML models
[144]	Thyroid cancer	Private data from 6 institutions	DL models outperformed FL models
[145]	Anal cancer	Private data from 3 institutions	Not available