Uneingeschränkter Zugang

Using explainable deep learning in da Vinci Xi robot for tumor detection

Rohan Ibn Azad
Rohan Ibn Azad
, 
Subhas Mukhopadhyay
Subhas Mukhopadhyay
 und 
Mohsen Asadnia
Mohsen Asadnia
   | 25. Okt. 2021
International Journal on Smart Sensing and Intelligent Systems's Cover Image
Über diesen Artikel
Vorheriger Artikel
Nächster Artikel
Zitieren
Online veröffentlicht: 25. Okt. 2021
Seitenbereich: 1 - 16
Eingereicht: 29. Juli 2021
Akzeptiert: 01. Sept. 2021
DOI: https://doi.org/10.21307/ijssis-2021-017
Schlüsselwörter
© 2021 Rohan Ibn Azad et al., published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Figure 1:

VGG-16 network used for classification.
VGG-16 network used for classification.

Figure 2:

Object detection stages (Bochkovskiy et al., 2020).
Object detection stages (Bochkovskiy et al., 2020).

Figure 3:

Object detection stages of YOLOv3 and YOLOv4.
Object detection stages of YOLOv3 and YOLOv4.

Figure 4:

Cropped photo from the surgical video. Contains tumor, portions of kidney, portions of fatty tissue (Abaza, 2020a, b; P. N. U. Specialist, n.d.).
Cropped photo from the surgical video. Contains tumor, portions of kidney, portions of fatty tissue (Abaza, 2020a, b; P. N. U. Specialist, n.d.).

Figure 5:

Tumors cropped from Figure 3 (Abaza, 2020a, b; P. N. U. Specialist, n.d.).
Tumors cropped from Figure 3 (Abaza, 2020a, b; P. N. U. Specialist, n.d.).

Figure 6:

Tumor detection on windows machine on videos (A) fatty tissue, (B) fatty tissue, (C) non-cancerous tissue, (D) cancerous tissue, non-cancerous tissue, fatty tissue.
Tumor detection on windows machine on videos (A) fatty tissue, (B) fatty tissue, (C) non-cancerous tissue, (D) cancerous tissue, non-cancerous tissue, fatty tissue.

Figure 7:

(A) VGG-16 2nd dataset, lower lr, Dropout 0.5, callback (B) VGG-16 3rd dataset, lower lr callback.
(A) VGG-16 2nd dataset, lower lr, Dropout 0.5, callback (B) VGG-16 3rd dataset, lower lr callback.

Figure 8:

Comparison of VGG-16 with lower lr, dropout, callback (left column) and VGG-16 with lower lr, callback (right column) for 5th dataset (3 class classification). (A) One of the images from the test set, (B) Coarse heatmap for the image from lower lr, dropout, callback, (C) Heatmap for the image from lower lr, dropout, callback. On the right column it was not detected. (D) The same image from the test set, (E) Network was not able to detect the image, that is why it is purple (F) No heatmap got detected on the image.
Comparison of VGG-16 with lower lr, dropout, callback (left column) and VGG-16 with lower lr, callback (right column) for 5th dataset (3 class classification). (A) One of the images from the test set, (B) Coarse heatmap for the image from lower lr, dropout, callback, (C) Heatmap for the image from lower lr, dropout, callback. On the right column it was not detected. (D) The same image from the test set, (E) Network was not able to detect the image, that is why it is purple (F) No heatmap got detected on the image.

Figure 9:

Research methodology flowchart.
Research methodology flowchart.

Comparison with other studies.

MethodImage typeAI technique usedTotal images (TI)Evaluation metricValidation performance (VP)VPTI
Hadjiyski (2020)CT scansInception v34,200AUC86%0.02
Aubreville et al. (2020)Whole Slide ImagesRetinaNet with ResNet-5013,907F1 score79.1%0.01
Wang et al. (2018)Multi parametric MRIV-net79 cases in total. About 790 imagesAccuracy89.4%0.11
Chung et al. (2015)Multi parametric MRISVM with RD-CRF20 cases in total. About 200 imagesAccuracy59%0.29
Brunese et al. (2020)Chest X-rayVGG-169,326Accuracy98%0.01
Wu et al. (2021)Chest CT scanVGG-16 with segmentation3,855Sensitivity95%0.03
This studyLive partial robotic nephrectomyObject detection with VGG-16143Accuracy84%0.59

Result comparison.

Detection algorithmPrecisionRecallMean average precisionFrames per second
YOLOv3 on virtual machine0.880.620.758Not applicable
YOLOv4 on windows0.980.990.97421.4

Institutions that produced the videos.

SourceCountryState
Brigham and Women’s HospitalUSAMassachusetts
Seattle Science FoundationUSAWashington
Pacific Northwest Urology specialistUSAWashington
Vattikuti FoundationUSAMichigan
Urologic Surgeons of WashingtonUSAWashington

Train, validation, test division.

DatasetLabelTrainingValidationTest
1st datasetCancerous tissue3095
Non-cancerous tissue40139
Fatty tissue21106
2nd datasetCancerous tissue10595
Non-cancerous tissue105139
Fatty tissue105106
3rd datasetCancerous tissue15095
Non-cancerous tissue1502315
eISSN:
1178-5608
Sprache:
Englisch
Zeitrahmen der Veröffentlichung:
Volume Open
Fachgebiete der Zeitschrift:
Technik, Einführungen und Gesamtdarstellungen, andere
Zeitschrift RSS Feed