Table 2

Literature review summary of lymph node metastasis (LNM) prediction in CRC using deep learning, split by patient-level or individual LN assessment

Patient-level LN assessment
Study detailsData summaryModel architectureKey limitationSensitivitySpecificityAccuracyAUC
Bedrikovetski et al,16 2023, Australia1201 CT scans, CC, 500 scans with LN annotationsShared ResNet-50 for segmentation and MLP classificationNon-expert annotations0.9660.05237.4%0.542
Wan et al,18 2023, China610 MRI scans, RC, *limited to T1–2,
tumour annotations
LN level 2D and 3D ResNet with transfer learning and averaging predictions from logistic regressionNot using cross-validation1.0000.66073.0%0.790
Liu et al,21 2023, China282 CT scans, RC,
Tumour annotations
ResNet-50, clinical and radiomics features, ML models and nomogram classifierTest set of 570.9550.85789.5%0.942
Xie et al,22 2023, China391 CT scans, CRC, LN annotationsLN and slice level ResNet-18 with an attention mechanism and logistic regressionNo external validation0.73474.7%0.768
Ding et al,17 2020, China545 MRI scans, RC, LN annotationsFaster R-CNN with transfer learning, clinical features, nomogram classifier and bounding box153/183 in the test set had LNM0.920
Glaser et al,23 2020, Australia123 CT scans, CC,
595 individual LN annotations
DenseNet encoder-decoder with segmentation and MLP classifierTest set of 23 patients0.860
Individual LN assessment
Study detailsData summaryModel architectureKey limitationSensitivitySpecificityAccuracyAUC
Ozaki et al,25 2023, Japan3547 LNs on MRI, RC, *Includes patients who had chemoradiotherapyResNet-18 with transfer learning, logistic regressionUsed same pathology label for pretreatment/post-treatment0.963
Li et al,24 2021, China129 MRI scans, RCInception-v3 with transfer learning, logistic regressionOnly included 1–2 LNs per patient0.9530.95295.7%0.994
Li et al,29 2021, China3364 MRI scans, CRCLeNet/AlexNet with transfer learning, ML models and MLPUsed radiologist LNM diagnosis0.8000.80075.8%0.794
Li et al,30 2021, China3364 MRI scans, CRCAlexNet with transfer learning and skip connections, MLPUsed radiologist LNM diagnosis0.8730.87483.6%0.857
Ding et al,32 2019, China414 MRI scans, RCFaster R-CNN with transfer learning, MLPAll patients had LNM
Li et al,28 2018, China619 LNs on MRI, CRCInception-v3 with transfer learning, MLPUsed radiologist LNM diagnosis94.4%
Lu et al,31 2018, China414 MRI scans, RCFaster R-CNN with VVG16 and transfer learning, MLP and bounding boxAll patients had LNM0.912
  • Includes study details, a description of the data with the number of patients or number of LNs, image modality (MRI/CT), cancer location (CRC or RC or CC) and unique inclusion criteria marked with an asterisk. Also, a model architecture summary, a key limitation and model performance metrics: sensitivity, specificity, accuracy and AUC to three significant figures. The best result for each metric is indicated in bold.

  • AUC, area under the curve; CC, colon cancer; CRC, colorectal cancer; RC, rectal cancer.