Table 2.

The summary of the application of AI in the treatment of CRC.

Theme	Year	Subject	Model	Sample	Results	Ref
Surgical treatment	2020	AI-assisted surgery	CNN	300 videos in operation	Accuracy = 81.0%; 83.2% (phase; action)	(118)
	2019	Robot-assisted surgery	da Vinci	Analyze from 206 RACRS patients	RM = 99.3%; 89.6% LN = 16 ± 6; 16 ± 8 LRR = 3.8%; 9.5% (colon; rectal)	(119)
	2021	AI-assisted LCRS	da Vinci	Analyze 600 images in 32 videos	DC = 0.84	(120)
	2022	Evaluate short outcomes	Senhance	Review outcomes in 55 Senhance assisted LCRC patients	Ileocecal resection = 32.7% high anterior resection = 20% D3 dissection = 74.5%	(121)
	2020	Automatic recognition	CNN	Recognize 71 Lap-S videos	Accuracy = 91.9%	(122)
	2021	Liver segment resection	da Vinci Xi	Present a video in a 54-year-old male patient	Operative time = 205 min estimated blood loss = 310 mL	(123)
	2020	Operation analysis	AIRAM	Test 25 ICG curve patterns	Processing time = 48.03 s	(24)
Chemoradiotherapy	2019	Assess therapy effect	RF	Assess performance from 55 patients	AUC = 0.86	(117)
	2022	Predict PCR after nCRT	RAPIDS	Study 933 patients	AUC = 0.812; sensitivity = 0.888 specificity = 0.740; NPV = 0.929 PPV = 0.512	(124)
	2021	Assess therapy effect	FFN/LR/SVM	Study 226 LARC patients	Accuracy = 0.67–0.75% AUC = 0.76–0.83% positive = 67–74%; NPV = 70–78% sensitivity = 68–79% specificity = 66–75%	(125)
	2018	Predict nCRT effect	DNN	Study 95 patients	Accuracy = 80%	(126)
	2020	Predict PCR after nCRT	ANN	Analyze 270 LARC patients	VSR = 1.57 (CEA levels)	(127)
	2022	Predict nCRT effect	MSCNN	Assess 150 WSI	AUC = 0.9337; 0.9091 (Camelyon; MSKCC)	(128)
	2019	Predict CRT response	CNN	Study 51 RC patients	AUC = 0.83	(129)
	2019	Predict nCRT effect	LR	Study 136 RC patients	AUC = 0.751; 0.831; 0.873 sensitivity = 66%; 71%; 75% specificity = 87.22%; 86.11%; 91.67% (pre-nCRT; early; combined)	(130)
	2020	Predict PCR,TRG, and NAR	LR	Collect and classify 132 nCRT and TME patients	AUC = 0.66; 0.80; 0.80 (NAR; PCR; TAG)	(131)
	2021	Predict and treat nCRT response	CFs-SVM	Analysis 428 patients	AUC = 0.834; 0.854 (training; validation)	(132)
Targeted therapy	2022	Identify therapy targets	MCODE	Extract four gene expression profile from database	Identify 8,931 DEGs in CRC patients	(133)
	2022	Design CAD approach	RF/SVP/CNN	Scanning 1,443 approved drugs	CAD design approach target p53 for treatment	(134)
	2022	Monitoring gene expression and drug effect	MLP	Study CRC cells genes phenomics	Mean accuracy = 9.48%↑(single track VS MLP)	(135)
	2021	Medicine precision	ML	Study STNs of CRC	The model with novel event freesurvival has a greater prediction	(136)
	2019	Tumor target segmentation	CAC-SPP	Evaluate two segmentation of tumor targets	DSC = 0.78 ± 0.08; 0.85 ± 0.03	(137)

AIRAM, artificial intelligence based real-time analysis microperfusion; AUC, area under the carve; ANN, artificial neural network; CNN, convolutional neutral network; CAD, computer aided drug; CAC, cascaded atrous convolution; DC, dice coefficient; DNN, deep neural network; DEGs, differential expressed genes; DSC, dice similarity coefficient; FFN, feedforward neural network; ICG, indocyanine green; LN, lymph nodes; LRR, locoregional recurrence rate; LCRS, laparoscopic colorectal surgery; Lap-s, laparoscopic sigmoidoscopy; LR, logistic regression; LARC, local advanced rectal cancer; MSCNN, multi-scale convolutional neural network; MCODE, molecular complex detection algorithm; MLP, machine learning phenomics; ML, machine learning; nCRT, neoadjuvant chemoradiotherapy; NPV, negative predictive value; NAR, neoadjuvant rectal score; PCR, pathological complete response; PPV, positive predictive value; RACRS, robot-assisted colorectal surgery; RM, radical margins; RF, random forest; RAPIDS, radiopathomics integrated prediction system; SVM, support vector machine; STNs, signal transduction network; SPP, spatial pyramid pooling; TRG, tumor regression grade; TME, total mesorectal excision; VSR, variable sensitivity ratio.