Table 8 Image analysis
Art | Mdl | Dom | Subdomain | Variables | Output | System training | Validation | Statistical outcome |
|---|---|---|---|---|---|---|---|---|
[142] | ANN | CaP | Radiotherapy dose planning | Patient prostate contour points (anterior, posterior and 5 lateral) | Anterior, posterior and lateral beam | 12–68 patients record of radiotherapy treatment planning | Average asymmetry of ANN and acceptance by dosimetrists Small field prostate (n = 133) and for large field prostate (n = 64) | Average asymmetry of ANN 0.20% and acceptance by dosimetrists was 96% (small field prostate) and 88% for large field prostate |
[143] | ANN | CaP | Diagnosis of localised disease from TRUS | Pixel distribution and grey levels of the TRUS images | Benign, malignant with Gleason grading | 53 images of benign and malignant sample images from 5 patients | Compare to histology results of 500 pictures from 61 patients post RRP for localised disease in one centre | Sp 99%, Se 83%, true positive for isoechoic is 97% |
[144] | ANN LDA | CaP | progression post RPP | Prostate volume, PSA, Pathology morphometric variables LDA | Progression or no | Progression t2n0 post RRP, 228 patients from one centre | ROC, Sp, Se, LOO 39 cases | Ac 70%, Se 55%, Sp 85% |
[144] | ANN LVQ | CaP | progression post RPP | Prostate volume, PSA, Pathology morphometric variables | Progression or no | Progression t2n0 post RRP, 228 patients from one centre | ROC, Sp, Se, LOO 39 cases | Ac 90%, Se 95%, Sp 85%, |
[144] | ANN LVQPAK | CaP | progression post RPP | Prostate volume, PSA, Pathology morphometric variables | Progression or no | Progression t2n0 post RRP, 228 patients from one centre | ROC, Sp, Se, LOO 39 cases | Ac 83%, Se 85%, Sp 80% |
[144] | ANN MLFF-bp | CaP | progression post RPP | Prostate volume, PSA, Pathology morphometric variables | Progression or no | Progression t2n0 post RRP, 228 patients from one centre | ROC, Sp, Se, LOO 39 cases | Ac 76%, Se 73%, Sp 80% |
[145] | kNN | CaP | TRUS cancer image analysis | Image pixels segmented by tissue descriptor (spatial grey level dependence) | Predict cancer | Images of 202 patients with suspected CaP at one centre | 87 randomly selected patients Comparison to other classifiers and ROC | AUC 0.6 |
[146] | ANN | CaP | TRUS Image segmentation | Pixel’s colour values from TRUS images | TRUS image segmentation | 212 CaP TRUS data | Overlap measure (compared to expert segmented boundary) on 10 random images | 81% mean overlap measurement |
[147] | ANN | CaP | MRI cancer diagnosis | 256 MRSI spectra (resonance intensities at given PPM) | Cancer or benign | 5308 voxels of 18 patients with CaP in a retrospective study | 15% validation ROC Se, Sp | AUC 0.95, Se 50%, Sp 99% |
[147] | ANN | CaP | MRI cancer diagnosis | 256 MRSI spectra (resonance intensities at given PPM), peripheral and transition zone, periurethral and outside region | Cancer or benign | 5308 voxels of 18 patients with CaP in a retrospective study | 15% CV validation ROC Se, Sp | AUC 0.97, Se 62%, Sp 99% |
[148] | SVM | CaP | Diagnosis of cancer from pMRI images | Image segmentation then clustering voxels | Cancer or benign | 16 pMRI images with CaP | Correlation coefficients of voxel parameters | Mean accuracy of 84% |
[149] | ANN | Bca | Image histology analysis | Image histology analysis (measurements of the segmentation of nuclei and other features) | Benign and malignant | 141 randomly chosen cell images (30%) | 329 cell images (70%) ROC, Sp, Se | Sp 100%, Se 82%, PPV 96%, NPV 80%, Ac 88% |
[150] | FCM | Bca | Diagnosis of tumour | Bladder wall segmentation and tumour region extraction To detect bladder abnormalities, four volume-based morphological features: bent rate, shape index, wall thickness, and bent rate difference between the inner and outer surfaces | Bladder neoplasm | 16 Bladder tumour MRI images | Overlap Ratio (OR) | OR 86.3% |
[151] | ANN | Bca | Transitional cell cytology analysis | Cytology image analysis and pixel variations as variables | Cancer or benign | 16 cytology images | comparison to experts, × 2 test | 75% concordance with the experts |
[152] | ANN | Nlt | Spectroscopy stone analysis | Absorbance infra-red spectrum of 91 wave lengths | Stone composition | 160 and 57 stone mixtures | Predictive accuracy, root mean square error on 36 independent stone samples | Overall good predictive value |