Copyright ©The Author(s) 2022.
World J Gastrointest Oncol. Jan 15, 2022; 14(1): 124-152
Published online Jan 15, 2022. doi: 10.4251/wjgo.v14.i1.124
Table 1 Artificial intelligence in diagnosis of colorectal cancer
Type of study
No. of participants
Control and interventions
Case control studyYang et al[19], 2019241Depth-learning intelligent assistant diagnosis systemBy comparing the accuracy of different algorithms on MRI images of patients with CRC, the algorithms that were conducive to the diagnosis of CRC were definedT2-weighted imaging method had obvious advantages over other methods in differentiating CRC
Analytical researchLiu et al[20], 2011429SVMCompared the performance of new and old classification methods in colorectal polyps CAD systemSVM could help CAD system get excellent classification performance
ReviewRegge et al[21], 2013NACAD systemNACAD system helped radiologists diagnose CRC with visual markers
Case control studySummers et al[22], 2008104CAD systemThe sensitivity of adenoma was measured by CAD system and compared with previous studiesCAD system had high accuracy in detecting and distinguishing adenoma
Descriptive researchChowdhury et al[23], 200853CAD-CTC systemThe sensitivity of CAD-CTC system and manual CTC was compared through the image data of 53 patientsCAD-CTC system could effectively identify polyps and cancers with clinical significance in CT images
Case control studyNappi et al[24], 2018196ResNetsBased on the clinical data of 196 patients, the classification performance of different models in distinguishing masses from normal colonic anatomy was comparedResNets solved the practical problem of how to optimize the performance of DL
Case control studyTaylor et al[25], 200824CAD systemThe effectiveness of CAD system in detecting tumors was tested using the clinical data of 24 patientsCAD could effectively detect flat carcinoma by tumor morphology
Case control studySummers et al[26], 2010394CAD-CTC systemThe CTC data sets of 394 patients were trained in CAD system. It was confirmed that the experimental group could reduce the missed diagnosis rate of cancerCAD-CTC system used advanced image processing and ML to reduce the occurrence of FP results
Case control studyLee et al[27], 201165CAD systemThe CTC data sets of patient polyps were divided into a training data set and a test data set to compare the detection performance of CAD systemCAD system included colon wall segmentation, polyp specific volume filter, cluster size counting and thresholding, which had high detection performance of polyps and cancer tissue
Case control studyNappi et al[28], 2015154DCNNThe clinical data were divided into a training data set and a test data set to compare the polyp detection performance of multiple classifiersDCNN could greatly improve the accuracy of automatic detection of polyps in CTC
Case control studyNäppi et al[29], 200514CAD systemThe clinical data of 14 patients were used to test the effect of different staining methods on the effectiveness of polyp detectionCAD system helped to improve the ability to detect polyps in CTC
Case control studyvan Wijk et al[30], 201084CAD-CTC systemThe polyp detection performance of different classification methods was tested through the clinical data of 84 patientsThe sensitivity of the CAD-CTC system to distinguish polyps over 6 mm was very high
Case control studyKim et al[31], 200735CAD systemThe sensitivity of CAD polyp detection was tested using colonoscopy data of 35 patientsCAD system helped to distinguish polyps and cancer tissue larger than or equal to 6 mm
Case control studyNappi et al[32], 2017101CADe systemThe polyp detection accuracy of novel and old CADe systems was compared by colonoscopy data of 101 patientsCADe system could improve the accuracy of detecting serrated polyps or cancer tissues
Case control studyMa et al[33], 2020681Portal venous phase timing algorithmTraining through 479 CT scan data sets; 202 CT scans were used for retrospective analysis and algorithm development and verificationIt was helpful to quantitatively describe the characteristics of tumor enhancement
Case control studySoomro et al[34], 2018123D fully convolutional neural networksThe effects of polyp segmentation and recognition of different models were compared using MRI data of 12 patients3D fully convolutional neural networks provided a more accurate segmentation result of colon MRI
Case control studySoomro et al[35], 201943DL43 patients with CRC were evaluated by MRI. The data set was divided into 30 volumes for training and 13 volumes for testingDL achieved better performance in colorectal tumor segmentation in volumetric MRI
Retrospective studyWang et al[36], 2020240Faster R-CNNThe Faster R-CNN was trained using pelvic MRI images to establish an AI platform. The diagnosis results of AI platform were compared with those of senior radiologistsIt was highly feasible to segment the circumcision positive margin with Faster R-CNN in MRI image of rectal cancer
Retrospective studyWu et al[37], 2021183Faster R-CNNThe MRI data of 183 patients were collected as training objects. The platform was constructed using Faster R-CNN. The diagnostic accuracy was compared with that of radiologistsAI could effectively predict the T stage of rectal cancer
Case control studyJoshi et al[38], 201010Non-parametric mixture modelCompared the accuracy of the algorithm and expert conclusions through the patient's MRI imagesThe algorithm could be used to distinguish T3 and T4 tumors accurately
Case control studyShiraishi et al[40], 2020314CNNThe prognostic significance was evaluated by CNN based on the expression of tumor markers in 314 patientsCNN could help to evaluate the diagnosis and prognosis of tumor markers
Case control studyPham[41], 2017NADLNADL could reduce training time and improve classification rate
Case control studyTiwari[42], 201810CNNCNN was used to compare the accuracy of image classification methods for seven different tissue typesCNN determined the most suitable color for cancer tissue classification (HSV color space) by classifying tissues in different color spaces
Case control studySirinukunwattana et al[43], 2016100SC-CNNThrough the comparative evaluation on the image data set of 100 cases of CRC, SC-CNN was helpful to the quantitative analysis of tissue componentsSC-CNN can help to predict the nuclear class tags more accurately
Case control studyKoohababni et al[44], 2018NADLNADL could combine the probability maps of a single nucleus to generate the final image, so as to improve the diagnostic performance of complex colorectal adenocarcinoma datasets
Case control studyZhang et al[45], 2018NAFaster R-CNNNAFaster R-CNN provided quantitative analysis of tissue composition in pathological practice
Case control studyXu et al[46], 20161376DCNNCompared the classification effects of AI and manual methods on the same pathological image datasetDCNN can help to improve the accuracy of differentiation between epithelial and mesenchymal regions in digital tumor tissue microarray
Retrospective studyChen et al[47], 201785Deep contour-aware networkThe classification performance of different segmentation methods on the same pathological image dataset was comparedOutput accurate probability map of gland cells, draw clear outline to separate the originally gathered cells, and further improve the segmentation performance
Case control studyYoshida et al[48], 20171328An automated image analysis systemThe classification results of the same dataset by human pathologists and electronic pathologists were comparedCompared with manual classification, the system had higher classification accuracy
Retrospective studySaito et al[49], 2013NACAD systemNACAD system could be used for quality control, double check diagnosis, and prevention of missed diagnosis of cancer
Descriptive researchJin et al[50], 2019NAAINAAI accelerated the transformation of pathology to quantitative direction, and provided annotation storage, sharing, and visualization services
Case control studyQaiser et al[51], 201975CNNThe segmentation and recognition effects of different methods on the same pathological dataset were comparedCNN and PHPs can more accurately and quickly distinguish tumor regions from normal regions by simulating the atypical characteristics of tumor nuclei
Retrospective studyZhou et al[53], 2020120DCNNIn the man-machine competition of 120 images, the accuracy of AI and endoscopists was comparedDCNN helped to establish an objective and stable bowel preparation system
Case control studyde Almeida et al[54], 2019NACNNNACNN improved the accuracy of polyp segmentation. It can help to automatically increase the sample number of medical image analysis dataset
Case control studyTaha et al[56], 201715DLThe effectiveness of the DL method for identifying polyps in colonoscopy images was verified on the public databaseIn the early screening of CRC, it was better than other single models
Case control studyYao et al[57], 2019NADLNAA DL algorithm in HSV color space was designed to effectively improve the accuracy of diagnosis and reduce the cost
Case control studyBravo et al[59], 2018NASupervised learning modelNASupervised learning model could help to detect polyps more than 5 mm automatically with high accuracy
Reviewde Lange et al[60], 2018NACAD systemNACAD system could eliminate the leakage rate of polyps, thus avoiding polyps from developing into CRC
Case control studyMahmood et al[61], 2018NACAD systemNACAD system combined with depth map could more accurately identify polyps or early cancer tissue
Retrospective studyMo et al[62], 201816DLCompared the performance of multiple algorithms in the same datasetDL was in the leading position in many aspects such as the performance of evolutionary algorithm, and was an effective clinical method
Case control studyZhu et al[63], 201050CAD systemThrough the database of 50 patients, the performance differences of different segmentation strategies were comparedInitial polyp candidates could greatly facilitate the FP reduction process of CAD system
Case control studyKomeda et al[64], 20171200CNN-CAD systemThe efficiency of CNN-CAD system was evaluated by maintaining cross validation for 10 timesCNN-CAD system can quickly diagnose colorectal polyp classification
Retrospective studyZhang et al[65], 201818CNN-CAD systemThrough the video of 18 cases of colonoscopy, the efficiency of polyp detection between CNN-CAD system and existing methods was comparedCNN-CAD system can reduce the chance of missed diagnosis of polyps
Case control studyZhu et al[66], 2019357CNNThe diagnostic performance of CNN was trained, fine-tuned, and evaluated using endoscopic data of 357 patients, and compared with that of manual diagnosisThe sensitivity of CNN optical diagnosis is higher than that of endoscopy, but the specificity is lower than that of endoscopy
Retrospective studyAkbari et al[67], 2018300FCNThe polyp segmentation method based on CNN was evaluated using CVC ColonDB databaseFCN proposed a new method of image block selection and the probability map was processed effectively
Retrospective studyYu et al[68], 2017NA3D-FCNNA3D-FCN could learn representative spatiotemporal features, and it had strong recognition ability
Case control studyYamada et al[69], 20194395AI The AI system was trained through a large amount of data to make it sufficient to detect missed non polypoid lesions with high accuracyAI could automatically detect the early features of CRC and improve the early detection rate of CRC
Retrospective studyLund et al[71], 201920DLPolyp video dataset was used as training data. At the same time, a 5-fold cross validation method was used to evaluate the accuracy of the systemDL could improve the network training efficiency of polyp detection accuracy
Meta-analysisTakamaru et al[73], 2020NAEndocytoscopyNAAI combined with endocytoscopy could greatly improve the efficiency of optical biopsy of CRC
ReviewDjinbachian et al[76], 2019NAAINAThe sensitivity of optical diagnosis based on AI could be comparable to that of experienced endoscopists
Retrospective studyKudo et al[77], 201969142EndoBRAINA retrospective comparative analysis was performed between EndoBRAIN and 30 endoscopists on the diagnostic performance of endoscopic images in the same datasetIn the image of color cell endoscopy, EndoBRAIN could distinguish between tumor and non-tumor lesions accurately
Retrospective studyMahmood et al[78], 2018NACRFNACRF estimated the depth of the colonoscopy image and reconstructed the surface structure of the colon
Case control studyJian et al[81], 20182772FCNQuantitative comparison of manual and AI segmentation results of 2772 cases of CRC in MRI imagesFCN was helpful for accurate segmentation of colorectal tumors
Case control studySivaganesan[82], 201620RNN-ALGAIn the same database, milestone algorithms such as graph cut and level set were compared with RNN-ALGA algorithmRNN-ALGA is suitable for abdominal slice of CT image, which can improve the accuracy and time efficiency of structure segmentation
Case control studyGayathri et al[83], 2015NANNNANN can help to remove the colonic effusion and obtain the ideal colon segmentation effect
Retrospective studyTherrien et al[84], 2018NASVM, CNNNAUsing multiple datasets to train SVM and CNN could more accurately distinguish CRC staining tissue than single dataset
Case control studySun et al[85], 2019NAMLNAML increased the chance of recognizing tumor bud by narrowing the region, thus providing effective tissue classification
Case control studyShi et al[86], 2010NADS-STMNADS-STM could reduce the cost of diagnosis
Case control studySu et al[87], 2012212MVMTMThe training set included 124 cases. The validation set included 88 cases. Comparedthe diagnostic efficiency of different methods for CRCCompared with the traditional ML method, MVMTM has the advantages of low cost
Case control studyKunhoth et al[88], 201780Multispectral image acquisition systemA group of 20 samples were selected from 4 different types of colorectal cells. Compared the accuracy of different feature extraction methodsThe database developed by this system had high classification accuracy
Case control studyWang et al[89], 20181290DLThrough the data of 1290 patients, an AI algorithm for real-time polyp detection was developed and verifiedCompared with ML, DL could detect polyps in real time and reduce the cost
Meta-analysisBarua et al[90], 2021NAAINAAI based polyp detection system could increase the detection of small non-progressive adenomas and polyps
Randomized controlled study Gong et al[91], 2020704ENDOANGEL system704 patients were randomly assigned to use the ENDOANGEL system for colonoscopy or unaided (control) colonoscopy to compare the efficiency of ENDOANGEL system with conventional colonoscopyThe system significantly improved the detection rate of adenoma in colonoscopy
Meta-analysisLui et al[92], 2020NAAINAAI system could improve the detection rate of adenoma and reduce the missed lesions in real-time colonoscopy
Case control studyRodriguez-Diaz et al[93], 2011134A diagnostic algorithm with ESS80 patients were randomly assigned to the training set, and the remaining 54 patients were assigned to the test set for prospective verification by the new algorithmThe algorithm with ESS reduced the risk and cost of biopsy, avoided the removal of non-neoplastic polyps, and reduced the operation time
Case control studyKondepati et al[94], 200737ANNThe tumor recognition accuracy of different algorithms was compared by collecting the spectra of cancer tissue and normal tissueThe spectrum was divided into cancer tissue group and normal tissue group by ANN, and the accuracy was 89%
Case control studyAngermann et al[95], 2016NAALNAAL helped to realize real-time detection and distinguish between polyps and cancer tissues
Case control studyAyling et al[96], 2019619ColonFlagTMThrough the clinical data of 619 patients, the performance of different systems in detecting CRC and high adenoma was comparedColonFlagTM could help special patients establish an appropriate safety net
Meta-analysisTian et al[97], 20204560EPETen randomized controlled trials were included and 4560 participants were included for meta-analysisEPE could guide the intestinal preparation of patients undergoing colonoscopy, and improve the detection rate of polyps, adenomas, and sessile serrated adenomas
Retrospective studyJaved et al[98], 2018NAQSLNAThe prevalent communities found by QSL represented different tissue phenotypes with biological significance
Case control studyWang et al[99], 2019328ANNDifferent diagnostic models were established by back propagation and other methods, and the performance of each model was evaluated by cross validation testANN combined with gene expression profile data could improve the diagnosis mode of CRC
Case control studyBattista et al[100], 2019345ANNThe diagnostic performance and FP of the new model were measured in the experimental group (patients with CRC) and the control group (patients with good health)ANN could help to establish an easily available, low-cost mathematical tool for CRC screening
ReviewZhang et al[101], 2021NAMLNAML based on cell-free DNA and microbiome data helped diagnose CRC
Case control studyWang et al[102], 20219631DCNNThe diagnostic accuracy of AI tools and experienced expert pathologists was compared through the same databaseA novel strategy for clinic CRC diagnosis using weakly labeled pathological whole-slide image patches based on DCNN
ReviewJones et al[103], 2021NAAINAElectronic health record type data combined with AI could help diagnose early cancer
Case control studyLorenzovici et al[104], 202133A computer aided diagnosis systemThe accuracy of the system in diagnosing CRC was tested through a dataset of 33 patientsThe system used ML to improve the accuracy of CRC diagnosis
Review and Meta-analysisXu et al[105], 2021NACNNNAThrough the comparative study of online database, CNN system had good diagnostic performance for CRC
Case control studyÖztürk et al[106], 2021NACNNNACNN was the most successful method that could effectively classify gastrointestinal image datasets with a small amount of labeled data
ReviewEchle et al[107], 2021NADLNADL could directly extract the hidden information from the conventional histological images of cancer, so as to provide potential clinical information
Table 2 Artificial intelligence in treatment of colorectal cancer
Type of study
Retrospective studyPassi et al[109], 2015DSS systemDSS system used follow-up data as a knowledge source to generate appropriate follow-up recommendations for patients receiving treatment
Retrospective studyLee et al[110], 2018Watson for OncologyWatson for Oncology could provide evidence-based treatment advice for oncologists
Retrospective studySiddiqi et al[111], 2008MATCH systemMATCH system could provide hundreds of data samples to help doctors choose the most personalized treatment plan
Retrospective studyLi et al[112], 2018 NanorobotNanorobots were relatively safe and immune inert. DNA nanorobots might represent a strategy for precise drug delivery in cancer treatment
Experimental study Felfoul et al[113], 2016NanorobotThe robot achieved an accurate effect of attacking cancer tumors
ReviewKoelzer et al[114], 2019MLThe combination of ML and computational pathology could inform the clinical choice and prognosis stratification of CRC patients
Retrospective studyLee et al[116], 2019Narrow-band imagingNarrow-band imaging helped doctors to predict the histology of colorectal polyps and estimate the depth of invasion
Meta-analysis, Case control studyIchimasa et al[117], 2018AIAI could reduce unnecessary surgery after endoscopic resection of stage T1 CRC without loss of lymph node metastasis
ReviewKirchberg et al[118], 2019Operation robotRobotic surgery had great potential, but it still needed high-quality evidence-based medicine
Experimental study Leonard et al[120], 2014Smart tissue autonomous robotSmart tissue autonomous robot was more accurate than surgeons using the most advanced robotic surgical system
Case control studyHuang et al[121], 2019Operation robotThe operation robot had the advantages of short operation time, low estimated bleeding, and fast recovery after operation
ReviewZheng et al[122], 2020Operation robotThere were some limitations, such as the disunity of technical standards and the excessive dependence on surgical robot equipment
ReviewMitsala et al[123], 2021Computer-assisted drug delivery techniquesThe technology could help to enhance the sensitivity and accuracy of targeted drugs
Case control studyAikemu et al[124], 2020AIAI provided personalized and novel evidence-based clinical treatment strategies for CRC
ReviewHamamoto et al[125], 2020AIAI provided a variety of new technologies for the treatment of CRC, such as surgical robots, drug localization technology, and various medical devices
ReviewPritzker[126], 2020AIAI could screen individual biomarkers for comprehensive and individualized treatment of colon cancer with low toxicity
Experimental studyDing et al[127], 2020AIThe drug dose optimization technology based on AI could achieve more accurate individualized treatment than traditional methods
Table 3 Artificial intelligence in prognosis evaluation of colorectal cancer
Type of study
Case control studyZhang et al[128], 2017Heterogeneous ensemble learning modelHeterogeneous ensemble learning model could use big data to identify high-risk groups of CRC patients
Retrospective studyMorgado et al[129], 2017Decision support systemDecision support system could evaluate the risk of CRC by processing incomplete, unknown, or even contradictory data
Case control studyAnand et al[131], 1999Intelligent hybrid systemEach AI technology produced a different set of important attributes. Intelligent hybrid system would be the trend of prognosis evaluation in the future
Case control studyGupta et al[132], 2019MLML could help to predict tumor stage and survival period
Case control studyLi et al[133], 2018MLCombining ML and database, clinicians might add race factor to evaluate prognosis
Case control studyBarsainya et al[134], 2018Decision tree classifierDecision tree classifier could predict recurrence and death according to various influencing factors
Cohort studyDimitriou et al[135], 2018MLA framework for accurate prognosis prediction of CRC based on ML datasets
Case control studyPopovici et al[136], 2017SVMThe accuracy of using SVM to distinguish CRC subtypes was very high
Experimental studyHoogendoorn et al[137], 2016AIAI helped doctors to extract useful predictors from non-coding medical records
Experimental studyKop et al[138], 2016MLThe combination of ML and electronic medical records could help early detection and intervention
Case control studyGeessink et al[139], 2015Supervised learningSupervised learning helped to predict the survival ability of tumor, so as to accurately stratify the prognosis of tumor patients
ReviewWright et al[140], 2014RFRF could reduce the workload of pathologists by automatically calculating the area ratio of each slide
Meta-analysisWang et al[141], 2019A two-stage ML modelCompared with the single-stage regression model, the two-stage model could obtain more accurate prediction results
Experimental study Oliveira et al[142], 2013CDSSCDSS based on cancer patients records and knowledge could provide support for surgeons
Meta-analysisLo et al[143], 2000CDSSCDSS could select the appropriate treatment from the aspects of curative effect, overall survival rate, and side effect rate
Case control studyHarrington et al[144], 2018MLML could be used to predict the risk of recurrence of colon polyps and cancer based on the pathological characteristics of medical records
Case control studyXie et al[145], 2018RF modelRF model helped to speculate the influencing factors of early CRC in China
Retrospective studyBokhorst et al[146], 2018DLDL helped reduce FP by detecting tumor bud
Cohort studyZhao et al[147], 2020DLThe method allowed objective and standardized application while reducing the workload of pathologists
Retrospective studySyafiandini et al[148], 2016DBMDBM helped to predict the survival time of cancer patients
Retrospective studyRoadknight et al[149], 2013MLML helped predict the prognosis of patients according to the immune status and other information
Case control studyCui et al[150], 2013SSLSSL improved the accuracy of predicting clinical results according to gene expression profile
Retrospective studyPark et al[151], 2014SSLSSL could improve the accuracy of predicting cancer recurrence
Retrospective studyDu et al[152], 2014Supervised learningSupervised learning could help to improve the accuracy of identifying cancer-related mutations
Case control studyChi et al[153], 2019Semi-supervised logistic regression methodSemi-supervised logistic regression method had better clinical prediction effect than supervised learning method
ReviewOng et al[154], 1997CARES systemCARES system helped early detection of cancer recurrence in high-risk patients
Case control studyReichling et al[155], 2020DGMateDGMate could judge the prognosis of tumor by detecting immunophenotype
Experimental study Chowdhury et al[156], 2011Crane algorithmCrane algorithm helped to describe the coordination of multiple genes and effectively predicted the metastasis of CRC
ReviewMohamad et al[157], 2019Nominal group techniqueNominal group technique was used in the content development of mobile app and the app used as a tool for CRC screening education
Retrospective studyHacking et al[158], 2020AIAI could improve the prognosis of patients by increasing the diagnostic accuracy of slide images