scholarly journals Assessing Generalisation Capabilities of CNN Models for Surgical Tool Classification

2021 ◽  
Vol 7 (2) ◽  
pp. 476-479
Author(s):  
Tamer Abdulbaki Alshirbaji ◽  
Nour Aldeen Jalal ◽  
Paul D. Docherty ◽  
Thomas Neumuth ◽  
Knut Moeller
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Visual Information ◽  
Comprehensive Evaluation ◽  
Context Aware ◽  
Surgical Tools ◽  
Surgical Tool ◽  
Crucial Component ◽  
Using Data ◽  
Laparoscopic Videos

Abstract Accurate recognition of surgical tools is a crucial component in the development of robust, context-aware systems. Recently, deep learning methods have been increasingly adopted to analyse laparoscopic videos. Existing work mainly leverages the ability of convolutional neural networks (CNNs) to model visual information of laparoscopic images. However, the performance was evaluated only on data belonging to the same dataset used for training. A more comprehensive evaluation of CNN performance on data from other datasets can provide a more rigorous assessment of the approaches. In this work, we investigate the generalisation capability of different CNN architectures to classify surgical tools in laparoscopic images recorded at different institutions. This research highlights the need to determine the effect of using data from different surgical sites on CNN generalisability. Experimental results imply that training a CNN model using data from multiple sites improves generalisability to new surgical locations.

Modular Neural Network for Learning Visual Features, Routes, and Operation Through Human Driving Data Toward Automatic Driving System

2020 ◽  
Vol 24 (3) ◽  
pp. 368-376
Author(s):  
Shun Otsubo ◽  
Yasutake Takahashi ◽  
Masaki Haruna ◽  
◽  
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Visual Information ◽  
Machine Learning Techniques ◽  
Driving System ◽  
Modular Neural Networks ◽  
Automatic Driving ◽  
Learning Module ◽  
Learning Techniques ◽  
Almost All

This paper proposes an automatic driving system based on a combination of modular neural networks processing human driving data. Research on automatic driving vehicles has been actively conducted in recent years. Machine learning techniques are often utilized to realize an automatic driving system capable of imitating human driving operations. Almost all of them adopt a large monolithic learning module, as typified by deep learning. However, it is inefficient to use a monolithic deep learning module to learn human driving operations (accelerating, braking, and steering) using the visual information obtained from a human driving a vehicle. We propose combining a series of modular neural networks that independently learn visual feature quantities, routes, and driving maneuvers from human driving data, thereby imitating human driving operations and efficiently learning a plurality of routes. This paper demonstrates the effectiveness of the proposed method through experiments using a small vehicle.

scholarly journals Deep learning for semantic segmentation of organs and tissues in laparoscopic surgery

2020 ◽  
Vol 6 (1) ◽  
Author(s):  
Paul Maria Scheikl ◽  
Stefan Laschewski ◽  
Anna Kisilenko ◽  
Tornike Davitashvili ◽  
Benjamin Müller ◽  
...  
Keyword(s):  
Neural Networks ◽  
Deep Learning ◽  
Laparoscopic Surgery ◽  
Semantic Segmentation ◽  
Cognitive Robotics ◽  
Loss Functions ◽  
Cognitive Systems ◽  
Context Aware ◽  
Segmentation Quality ◽  
And Training

AbstractSemantic segmentation of organs and tissue types is an important sub-problem in image based scene understanding for laparoscopic surgery and is a prerequisite for context-aware assistance and cognitive robotics. Deep Learning (DL) approaches are prominently applied to segmentation and tracking of laparoscopic instruments. This work compares different combinations of neural networks, loss functions, and training strategies in their application to semantic segmentation of different organs and tissue types in human laparoscopic images in order to investigate their applicability as components in cognitive systems. TernausNet-11 trained on Soft-Jaccard loss with a pretrained, trainable encoder performs best in regard to segmentation quality (78.31% mean Intersection over Union [IoU]) and inference time (28.07 ms) on a single GTX 1070 GPU.

scholarly journals Surgical Tool Classification in Laparoscopic Videos Using Convolutional Neural Network

2018 ◽  
Vol 4 (1) ◽  
pp. 407-410 ◽  
Author(s):  
Tamer Abdulbaki Alshirbaji ◽  
Nour Aldeen Jalal ◽  
Knut Möller
Keyword(s):  
Neural Network ◽  
Convolutional Neural Network ◽  
Classification Performance ◽  
Training Data ◽  
Negative Effects ◽  
Surgical Tools ◽  
Surgical Tool ◽  
Context Aware System ◽  
Source Of Information ◽  
Laparoscopic Videos

AbstractLaparoscopic videos are a very important source of information which is inherently available in minimally invasive surgeries. Detecting surgical tools based on that videos have gained increasing interest due to its importance in developing a context-aware system. Such system can provide guidance assistance to the surgical team and optimise the processes inside the operating room. Convolutional neural network is a robust method to learn discriminative visual features and classify objects. As it expects a uniform distribution of data over classes, it fails to identify classes which are under-presented in the training data. In this work, loss-sensitive learning approach and resampling techniques were applied to counter the negative effects of imbalanced laparoscopic data on training the CNN model. The obtained results showed improvement in the classification performance especially for detecting surgical tools which are shortly used in the procedure.

A deep learning spatial-temporal framework for detecting surgical tools in laparoscopic videos

2021 ◽  
Vol 68 ◽  
pp. 102801
Author(s):  
Tamer Abdulbaki Alshirbaji ◽  
Nour Aldeen Jalal ◽  
Paul D. Docherty ◽  
Thomas Neumuth ◽  
Knut Möller
Keyword(s):  
Deep Learning ◽  
Surgical Tools ◽  
Laparoscopic Videos

scholarly journals Liver CT Image Processing And Diagnosing Using Artificial Neural Networks And MATLAB

10.29007/3dj7 ◽  
2020 ◽  
Author(s):  
Nhat Nguyen Thanh Minh ◽  
Van Hoang Tien Tran
Keyword(s):  
Neural Networks ◽  
Image Processing ◽  
Deep Learning ◽  
Network Architecture ◽  
Active Contour ◽  
Grand Challenge ◽  
Liver Ct ◽  
Biomedical Image ◽  
Using Data ◽  
Care Problems

Segmentation is a grand challenge, and there are many contests are held around the world to solve this challenge, especially in the biomedical image. There are many solutions to solve this challenge have been published.Nowadays, neural networks, including deep learning is a powerful and state-of-the-art way to segment objects from the background. But to use deep learning effectively, besides design a good network architecture, the preparation of input data is also an important requirement. Active contour (another name: Snake) is a classical segmentation technique in image processing. But the accuracy of this technique is not as high as we need for health care problems, and soft techniques such as neural networks or deep learning can improve this problem. But in those researches, deep learning is supplied to change the parameters of the active contour algorithm.We propose a combination of two fields of solving segmentation problem, a classical one, and a modern: using data from active contour to be the input of deep learning. The images to be used in this research are human liver CT images.

Deep Learning in mHealth for Chronic Diseases: A Systematic Review (Preprint)

2021 ◽  
Author(s):  
Andreas Triantafyllidis ◽  
Haridimos Kondylakis ◽  
Dimitrios Katehakis ◽  
Angelina Kouroubali ◽  
Lefteris Koumakis ◽  
...  
Keyword(s):  
Systematic Review ◽  
Neural Networks ◽  
Cardiovascular Disease ◽  
Deep Learning ◽  
Chronic Diseases ◽  
Real Life ◽  
Bibliographic Databases ◽  
Target Disease ◽  
Using Data ◽  
Almost All

BACKGROUND Major chronic diseases such as cardiovascular disease, diabetes, and cancer impose a significant burden on people and the healthcare systems around the globe. Recently, Deep Learning (DL) has shown great potential towards the development of intelligent mobile health (mHealth) interventions for chronic diseases which could revolutionize the delivery of healthcare anytime-anywhere. OBJECTIVE To present a systematic review of studies which have used DL based on mHealth data for the diagnosis, prognosis, management, and treatment of major chronic diseases, and advance our understanding of the progress made in this rapidly developing field. METHODS We searched the bibliographic databases of Scopus and PubMed in order to identify papers with focus on the employment of DL algorithms using data captured from mobile devices (e.g., smartphones, smartwatches, and other wearable devices), and targeting cardiovascular disease, diabetes, or cancer. The identified studies were synthesized according to the target disease, the number of enrolled participants and their age, the study period, as well as the employed DL algorithm, the main DL outcome, the dataset used, the features selected, and the achieved performance. RESULTS 20 studies were included in the review. 7 DL studies (35%) targeted cardiovascular disease, 9 studies (45%) targeted diabetes, and 4 studies (20%) targeted cancer. The most common DL outcome was diagnosis of patient condition for the cardiovascular disease studies, prediction of blood glucose values for studies in diabetes, and early detection of cancer. The DL algorithms employed most were convolutional neural networks and recurrent neural networks. The performance of DL was found overall to be satisfactory reaching more than 84% accuracy in the majority of the studies. Almost all studies did not provide details on the explainability of DL outcomes. CONCLUSIONS The use of DL can facilitate the diagnosis, management and treatment of major chronic diseases through harnessing mHealth data. Prospective studies are now required to demonstrate the value of applied DL in real-life mHealth interventions.

Levenshtein Augmentation Improves Performance of SMILES Based Deep-Learning Synthesis Prediction

2020 ◽  
Author(s):  
Dean Sumner ◽  
Jiazhen He ◽  
Amol Thakkar ◽  
Ola Engkvist ◽  
Esben Jannik Bjerrum
Keyword(s):  
Neural Networks ◽  
Pattern Recognition ◽  
Deep Learning ◽  
Recurrent Neural Networks ◽  
Data Augmentation ◽  
State Of The Art ◽  
Sequence Similarity ◽  
Learning Models ◽  
Underlying Network

<p>SMILES randomization, a form of data augmentation, has previously been shown to increase the performance of deep learning models compared to non-augmented baselines. Here, we propose a novel data augmentation method we call “Levenshtein augmentation” which considers local SMILES sub-sequence similarity between reactants and their respective products when creating training pairs. The performance of Levenshtein augmentation was tested using two state of the art models - transformer and sequence-to-sequence based recurrent neural networks with attention. Levenshtein augmentation demonstrated an increase performance over non-augmented, and conventionally SMILES randomization augmented data when used for training of baseline models. Furthermore, Levenshtein augmentation seemingly results in what we define as <i>attentional gain </i>– an enhancement in the pattern recognition capabilities of the underlying network to molecular motifs.</p>

Sign in / Sign up

Export Citation Format

Share Document