A sequential search-space shrinking using CNN transfer learning and a Radon projection pool for medical image retrieval

Khatami, Amin; Babaie, Morteza; Tizhoosh, H.R.; Khosravi, Abbas; Nguyen, Thanh; Nahavandi, Saeid

doi:10.1016/j.eswa.2018.01.056

Cited by 101 publications

(45 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Anyway, the learning is transferable to the new task by finetuning the pretrained CNN: the feature representations learned for the original task are slightly adjusted for supporting the new task (in this case, an object detection task). These slight adjustments may be executed by a small training set [9,10], which is only available for the object detector.…”

Section: Convolutional Neural Network For Object Detectorsmentioning

confidence: 99%

A Dropout Technique Study for the Faster R-CNN Detectors With Pretrained Convolutional Neural Networks for Detecting Very Simple Objects That Can Be Masked

Romanuke¹

2018

VAM

View full text Add to dashboard Cite

One of the best object detection methods, the Faster R-CNN, uses a pretrained convolutional neural network allowing to train the detector on small training sets typical in the object detection practice. Convolutional networks are prevented from overfitting by inserting DropOut layers. An open question is whether the DropOut technique improves much the object Faster R-CNN detector accuracy. Therefore, the goal is to show how the DropOut technique influences on the object detector performance. An original image classification dataset for pretraining a convolutional neural network is CIFAR-10. An appropriate convolutional network architecture for classifying CIFAR-10 images has a 50 % DropOut layer inserted in-between two fully-connected layers. Object detection tasks used for training and testing the Faster R-CNN detector are of monochrome images wherein small black rectangles are to be detected. Despite such objects are very simple, they can be masked around some dark localities so that detection would not be easy. One detection task is to detect the black rectangles in suburb house frontal views. Another one is to detect the rectangles in office room views. The suburb view dataset is divided into a training set of 120 images and a testing set of 121 images, every entry with a black rectangle. The office view dataset is divided likewise into a training set of 115 images and a testing set of 100 images, every entry with a black rectangle. Performance of the detector is studied against three training parameters: bounding box overlap ratio for positive training samples, minimum anchor box size, and anchor box pyramid scale factor. The performance is meant by the number of detected objects along with the intersection-over-union. However, neither graphs for the summed intersection-over-unions, nor graphs for the number of detected objects show that the DropOut technique influences on the Faster R-CNN object detector performance. Even for letting miss a few objects and decreasing an accuracy threshold, this influence is not significant. Therefore, a pretrained convolutional neural network to be included into the Faster R-CNN object detector should not contain a DropOut layer, especially if the network is trained much longer with the DropOut layer.Key words: object detection, Faster R-CNN object detector, pretrained convolutional network, DropOut, monochrome image, training set, intersection-over-union, number of detected objects. DROPOUT LAYERS FOR PREVENTING OVERFITTINGOverfitting is a poor effect of an analysis over data that involves too complex modeling. In machine learning, overfitting is especially likely for small training data sets and, additionally, when training is performed overlong. To prevent overfitting, a technique named dropout (or, stylistically, DropOut) is used [1,2]. In convolutional neural networks (CNNs), DropOut is a layer, at which a part of nodes is randomly removed while training. The randomness of the remover is defined with a probability [2,3]. This probability is commonly equal to 0.5, unles...

show abstract

Section: Convolutional Neural Network For Object Detectorsmentioning

confidence: 99%

A Dropout Technique Study for the Faster R-CNN Detectors With Pretrained Convolutional Neural Networks for Detecting Very Simple Objects That Can Be Masked

Romanuke¹

2018

VAM

View full text Add to dashboard Cite

show abstract

“…It is much hard to manage and analyze after which are stored in the database. These stored medical images can"t be retrieve easily from large database, because its search space is much large [18]. For making efficient decision and diagnosis on medical field the authors of [19] presented an article for retrieving such medical images.…”

Section: A Retrieval Of Medical Imagementioning

confidence: 99%

A Study of Retrieval Methods of Multi-Dimensional Images in Different Domains

Garg¹

2018

ijacsa

View full text Add to dashboard Cite

Multiple amount of multi-dimensional images are designed and most of them are available on internet at free of cost. The 3D images include three characteristics namely width, height, and depth. The images which are created as 3D can describe the geometry in terms of 3D coordinates. These coordinates help to obtain the object from the image much easier and accurate. In this paper, we presented a review about the Multi-dimensional image retrieval. Multi-dimensional image retrieval is a process of extracting the relevant 2D or 3D images from the huge database. To perform image retrieval process on large database, several methods like text based, Content based, Annotation based, semantic based, and sketch based were used. The image retrieval techniques are mostly used in the fields like Digital library, medical, forensic science, and so on. A systematic literature review has been shown for image retrieval methods reported on 2010 to 2017. The aim of this article is to show the various concept and efforts of different authors on image retrieval technique.

show abstract

“…Imaging-based classification primarily identifies the object of interest to be malignant or benign. Most recently, deep learning (Lecun, Bengio, & Hinton, 2015) has gained great success in performing all three tasks (Affonso Carlos, Renato, & Marques, 2015;He, Zhang, Ren, & Sun, 2016;Khatami et al, 2018;Szegedy et al, 2015). Deep learning owes its success largely to the fact that its models are capable of learning and reproducing an extensive range of parameters from the layers.…”

Section: Introductionmentioning

confidence: 99%

A feature transfer enabled multi-task deep learning model on medical imaging

Gao

Yoon

et al. 2020

Expert Systems with Applications

View full text Add to dashboard Cite

Object detection, segmentation and classification are three common tasks in medical image analysis. Multi-task deep learning (MTL) tackles these three tasks jointly, which provides several advantages-saving computing time and resources and improving robustness against overfitting.However, existing multi-task deep models start with each task as an individual task and integrate parallelly conducted tasks at the end of the architecture with one cost function. Such architecture fails to take advantage of the combined power of the features from each individual task at an early stage of the training. In this research, we propose a new architecture, FT-MTL-Net, an MTL enabled by feature transferring. Traditional transfer learning deals with the same or similar task from different data sources (a.k.a. domain). The underlying assumption is that the knowledge gained from source domains may help the learning task on the target domain. Our proposed FT-MTL-Net utilizes the different tasks from the same domain. Considering features from the tasks are different views of the domain, the combined feature maps can be well exploited using knowledge from multiple views to enhance the generalizability. To evaluate the validity of the proposed approach, FT-MTL-Net is compared with models from literature including 8 classification models, 4 detection models and 3 segmentation models using a public full field digital mammogram dataset for breast cancer diagnosis. Experimental results show that the proposed FT-MTL-Net outperforms the competing models in classification and detection and has comparable results in segmentation.

show abstract

A sequential search-space shrinking using CNN transfer learning and a Radon projection pool for medical image retrieval

Cited by 101 publications

References 26 publications

A Dropout Technique Study for the Faster R-CNN Detectors With Pretrained Convolutional Neural Networks for Detecting Very Simple Objects That Can Be Masked

A Dropout Technique Study for the Faster R-CNN Detectors With Pretrained Convolutional Neural Networks for Detecting Very Simple Objects That Can Be Masked

A Study of Retrieval Methods of Multi-Dimensional Images in Different Domains

A feature transfer enabled multi-task deep learning model on medical imaging

Contact Info

Product

Resources

About