Table 1 Limitations and outcomes of previous work.
Author | Dataset | Dataset type | Technique | Outcomes | Limitations | ||
|---|---|---|---|---|---|---|---|
Data augmentation | Improvable Accuracy | Use of XAI | |||||
Chiang et al.47 | 105 healthy controls and 151 patients with calcium oxalate stones | Handcrafted features | ANN and DA | ANN achieved an accuracy of 89%, while DA achieved 75% | No | Yes | No |
Dussol et al.48 | 119 stone formers and 96 controls | Handcrafted features | ANN (LDA and MVDA) | ANN with LDA gives a high accuracy of 75.8% compared to ANN with MVDA (74.4%) | No | Yes | No |
Cauderella et al.49 | 80 patient’s data | Handcrafted features | ANN and LR | Comparing the performance of ANN with LR and then observing that ANN gives better accuracy of 88.8% compared to LR (67.5%) | No | Yes | No |
Kumar and Abhishek50 | Data from 1000 patients | Handcrafted features | LVQ, MLP, and RBF | The accuracy obtained by LVQ, MLP, and RBF is 84%, 92%, and 87%, respectively | No | Yes | No |
Ebrahimi and Mariano51 | KUB CT scan slides from 39 patients | Image-based | Image processing techniques and geometry principles | Detect kidney stones with an accuracy of 84.61% | No | Yes | No |
Kazemi and Mirroshandel52 | Numeric characteristics from 936 patients | Handcrafted features | Ensemble learning model | Obtained an accuracy of 97.1% | No | Yes | No |
Li and Elliot53 | 1874 CT KUB reports | Handcrafted features | NLP | An overall accuracy of 85% was attained by applying NLP to CT KUB reports | No | Yes | No |
De Perrot et al.54 | 416 patient data | Handcrafted features | ML model | Using a machine learning model results in an overall accuracy of 85.1% | No | Yes | No |
Kahani et al.55 | KUB x-ray images | Image-based | LASSO with ML classifiers | Obtained an accuracy of 96% | No | Yes | No |
Jungmann et al.56 | 1714 LDCT images | Image-based | NLP | Applying NLP to 1714 LCDT images achieves an overall accuracy of 72% | No | Yes | No |
Annameti Rohith et al.57 | 114 ultrasound images | Image-based | Median and rank filters | When applied to 114 ultrasound images, the median filter gives an overall high accuracy of 86.4% compared to the rank filter (82.2%) | No | Yes | No |
Suresh and Abhishek58 | KUB ultrasound images | Image-based | Image processing techniques | The proposed model gives an accuracy of 92.57% for stone detection | No | Yes | No |
Jendenber et al.59 | NCCT images of 341 patients containing a distal ureteral stone, phlebolith, or both | Image-based | CNN | CNN differentiated stones and phlebolith with 92% accuracy | No | Yes | No |
Cui et al.60 | 625 CT images | Image-based | DL and threshold-based model | Achieved an accuracy of 90.30% | No | Yes | No |
Yildirim et al.61 | 1799 coronal CT images | Image-based | XResNet-50 | Using CT images, XResNet-50 demonstrated an accuracy of 96.82% | No | Yes | No |
Tsitsiflis et al.62 | Medical data of 716 patients | Feature-based data | ANN | Achieved a testing accuracy of 81.43% | No | Yes | No |
Valencia et al.63 | CT scans of around 40 patients | Image-based | Image processing | Achieved an accuracy of 92.5% for stone detection | No | Yes | No |
