Table 1 Data-driven precision diagnosis in digestive diseases based on radiomics
From: Data-driven decision-making for precision diagnosis of digestive diseases
First author, year | Disease | n | Data source and specific task | ML method | Diagnostic performance | Refs. |
|---|---|---|---|---|---|---|
Liu, 2022 | GIST | 106 | Abdominal CT image; VOI segmentation, image normalization and feature extraction | GLM/LASSO | Accuracy: 80.8% | [47] |
Kim, 2021 | CRC | 502 | Abdominal CT image; ROI segmentation, feature extraction | CNN/Transfer Learning | Sensitivity: 81.82% | [48] |
Ma, 2017 | GC | 40 | Abdominal CT image; VOI segmentation, feature extraction | LASSO | Accuracy: 81.43% | [49] |
Yasar, 2019 | GC | 10 | Endoscopic image; image-based segmentation | Clustering | Accuracy: 96.33% | [51] |
Li, 2021 | Crohn disease | 167 | Abdominal CT enterography; VOI segmentation, feature extraction and selection | LASSO | AUC: 0.816 (95%CI, 0.706–0.926) | [52] |
Yuan, 2022 | CRC | 140 | Abdominal CT image; manual contouring, image-based ResNet-3D base neuron | ResNet3D/SVM | AUC: 0.922 (95%CI, 0.912–0.944) | [53] |
Wu, 2022 | Hepatic cystic echinococcosis | 967 | Abdominal ultrasound image; artificial marker repair and ROI extraction, image-based classification | DCNN | Accuracy: 90.6% | [54] |
Kundu, 2020 | Multi-disease detection | 50 | WCE image; image ROI separation, probability density function | LDA/Hierarchical SVM | Accuracy: 97.39% | [55] |
Klang, 2020 | Crohn disease | 49 | WCE image; image-based classification | CNN | Accuracy: 95.4–96.7% | [56] |
Dmitriev, 2020 | Pancreatic cystic lesions | 134 | Abdominal CT image; graph-based segmentation | RF/CNN | Accuracy: 91.7% | [57] |
Meng, 2022 | Crohn disease | 235 | Abdominal CT enterography; image ROI separation, patch-based classification | 3D DCNN | AUC: 0.808–0.839 | [58] |
Wang, 2023 | GHAC | 216 | Abdominal CT image; image ROI segmentation and radiomics feature extraction | LASSO | AUC: 0.731–0.942 | [59] |
Shi, 2023 | PMME | 122 | Chest CT image; image resampling, tumor segmentation and feature extraction | LASSO | AUC: 0.906–0.975 | [60] |
Zhou, 2023 | Crohn disease | 316 | CT enterography; VAT features extraction | PCA/LASSO/3D-CNN | AUC: 0.775 (95%CI, 0.683–0.868) | [61] |
Sun, 2019 | GC | 100 | Abdominal CT image; ROI segmentation and radiomics feature extraction | LASSO | AUC: 0.903 | [62] |
Lonseko, 2023 | GI lesion | 4880 | GI endoscopic image; gastrointestinal lesion segmentation | GANs/CNN | Precision: 91.72% ± 4.05% | [63] |
Jia, 2023 | GIST | 151 | Abdominal CT image/EUS image; image segmentation, image normalization, and feature extraction | LASSO | AUC: 0.766–0.866 | [64] |
Guo, 2022 | CRC | 360 | Abdominal imaging examination data; ROI segmentation and feature extraction | CNN/K-means clustering | AUC: 0.950 | [65] |
Du, 2023 | gastric neoplasms | 3449 | WL and WM endoscopy image and video; ROI segmentation and feature extraction | CNN | Accuracy: 90.0% | [66] |
Tang, 2023 | GI tract diseases | 1645 | GI endoscopic image; classification and segmentation | TransMT-Net | Accuracy: 96.9% | [67] |
Gong, 2023 | gastric neoplasms | 8993 | GI endoscopic image; semantic segmentation | U-Net + + /CNN | Accuracy: 95.6% | [68] |
Yang, 2023 | Intestinal Metaplasia Gastritis Atrophy | 21,420 | Gastric endoscopic image; localization, patch-based classification | LAG/DTL | Accuracy: 97.1–99.2% | [69] |
Ding, 2023 | GI lesion | 2565 | Capsule endoscopy image and video; image-based classification | CNN/CRNN | Accuracy: 79.2–97.5% | [70] |
Muniz, 2023 | CRC | 71 | Micro-FTIR absorbance HSI from biopsy tissue; localization, voxel-based classification | FCNN/linear SVM | Accuracy: 96–99% | [71] |
Du, 2023 | GC | 1273 | Gastroscopic image; segmentation, co-spatial attention and channel attention | CSA–CA–TB–ResUnet | Accuracy: 91.2% | [72] |
Yuan, 2023 | GC | 4315 | Tongue image; patch-based classification | KNN/SVM/DT/APINet/TransFG/DeepLabV3 + | AUC: 0.830–0.920 | [73] |
Faust, 2023 | Celiac Disease | 96 | Duodenitis biopsy image; CLAHE, feature extraction | SVM/KNN/DT | Accuracy: 98.5–98.6% | [74] |
Kim, 2023 | CRC | 889 | CRC histopathologic slide; patch extraction and normalization, patch-based classification | CNN | Accuracy: 95.5% | [75] |
Abdelrahim, 2023 | Barrett's neoplasia | 270 | Gastroscopy image and video; image-based classification | CNN | Accuracy: 92.0–94.7% | [76] |
Fockens, 2023 | Barrett's neoplasia | 4920 | WL endoscopy image; segmentation, image-based classification | EfficientNet‐Lite1/MobileNetV2 DeepLabV3 + | Sensitivity: 84–100% | [77] |
Zhang, 2023 | gastrointestinal disorders | 315,767 | Gastroscopy image and video; localization, video-based classification | DCNN | Accuracy: 73.1–85.2% | [78] |
Zhou, 2023 | gastric polyps/gastric ulcers/gastric erosions | 227 | Gastroscopic image; feature extraction, feature fusion, image-based classification | GoogLeNet/ResNet/ResNeXt/SVM/RF | Accuracy: 81.7–82.5% | [79] |
Fan, 2023 | UC | 332 | Endoscopic image and video; feature extraction, image-based classification | CNN | Accuracy: 86.54% | [80] |
Faghani, 2022 | Barrett's esophagus | 542 | Esophagus histology slide; image-based classification | CNN | Sensitivity: 90–100% | [81] |
Yang, 2022 | upper GI diseases | 9403 | GI endoscopic image; image-based classification | VGG-11/ResNet50/DenseNet121 | Accuracy: 91.8% | [82] |
Yuan, 2022 | ESCC | 685 | GI endoscopic image; feature extraction, patch-based classification | DCNN | Accuracy: 89.8–91.3% | [83] |
Luo, 2022 | CAG | 4005 | GI WL image; image-based classification | CNN | Accuracy: 85.4–91.6% | [84] |