A Multiple-instance Learning Approach for the Assessment of Gallbladder Vascularity from Laparoscopic Images

Loukas, Constantinos; Gazis, Athanasios; Schizas, Dimitrios

doi:10.5220/0010762500003123

Cited by 1 publication

(6 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More background information about the MIL algorithms can be found in 31 . Regarding the feature extraction required by the baseline techniques, we employed an initial set of 707 colour, texture and statistical features described in our prior work 22 . After principal component analysis the feature vector was reduced to about 210 dimensions (depending on the training fold).…”

Section: Methodsmentioning

confidence: 99%

“…Part of the current dataset has been employed in the past for similar purposes 18 ,. 22 The outlined GB regions were also annotated by two expert surgeons from our institution with respect to the vascularity level of the GB wall: low ( L ) and high ( H ). Label H denotes presence of prominent superficial vessels whereas label L denotes absence of, or minor, vessels and significant fat coverage.…”

Section: Methodsmentioning

confidence: 99%

“…This approach limits the manual labelling process only to the GB regions, proving significant reduction in the annotation effort. However in 22…”

Section: Introductionmentioning

confidence: 99%

“…Recently, this limitation was alleviated via multiple instance learning (MIL). 22 Compared to singleinstance learning (SIL), in MIL algorithms the labels are assigned only to the GB regions (bags) whereas the labels of the patches (instances) are unknown. This approach limits the manual labelling process only to the GB regions, proving significant reduction in the annotation effort.…”

Section: Introductionmentioning

confidence: 99%

“…A significant challenge was the manual effort required to annotate patches for CNN training. Recently, this limitation was alleviated via multiple instance learning (MIL) 22 . Compared to single‐instance learning (SIL), in MIL algorithms the labels are assigned only to the GB regions (bags) whereas the labels of the patches (instances) are unknown.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Multiple instance convolutional neural network for gallbladder assessment from laparoscopic images

Loukas

Gazis

Schizas

2022

Robotics Computer Surgery

Self Cite

View full text Add to dashboard Cite

Background We present an artificial intelligence framework for vascularity classification of the gallbladder (GB) wall from intraoperative images of laparoscopic cholecystectomy (LC). Methods A two‐stage Multiple Instance Convolutional Neural Network is proposed. First, a convolutional autoencoder is trained to extract feature representations from 4585 patches of GB images. The second model includes a multi‐instance encoder that fetches random patches from a GB region and outputs an equal number of embeddings that feed a multi‐input classification module, which employs pooling and self‐attention mechanisms, to perform prediction. Results The evaluation was performed on 234 GB images of low and high vascularity from 68 LC videos. Thorough comparison with various state‐of‐the‐art multi‐instance and single‐instance learning algorithms was performed for two experimental tasks: image‐ and video‐level classification. The proposed framework shows the best performance with accuracy 92.6%–93.2% and F1 93.5%–93.9%, close to the agreement of two expert evaluators (94%). Conclusions The proposed technique provides a novel approach to classify LC operations with respect to the vascular pattern of the GB wall.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

“…This approach limits the manual labelling process only to the GB regions, proving significant reduction in the annotation effort. However in 22…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%