Weakly supervised object detection with 2D and 3D regression neural networks

Dubost, Florian; Adams, Hieab; Yilmaz, Pinar; Bortsova, Gerda; Tulder, Gijs van; Ikram, M. Arfan; Niessen, W.J.; Vernooij, Meike W.; Bruijne, Marleen de

doi:10.1016/j.media.2020.101767

Weakly supervised object detection with 2D and 3D regression neural networks

Title

Weakly supervised object detection with 2D and 3D regression neural networks

Author

Dubost, Florian (Erasmus MC)
Adams, Hieab (Erasmus MC)
Yilmaz, Pinar (Erasmus MC)
Bortsova, Gerda (Erasmus MC)
Tulder, Gijs van (Erasmus MC)
Ikram, M. Arfan (Erasmus MC)
Niessen, W.J. (TU Delft ImPhys/Medical Imaging; TU Delft ImPhys/Computational Imaging; Erasmus MC)
Vernooij, Meike W. (Erasmus MC)
Bruijne, Marleen de (Erasmus MC; University of Copenhagen)

Date

2020

Abstract

Finding automatically multiple lesions in large images is a common problem in medical image analysis. Solving this problem can be challenging if, during optimization, the automated method cannot access information about the location of the lesions nor is given single examples of the lesions. We propose a new weakly supervised detection method using neural networks, that computes attention maps revealing the locations of brain lesions. These attention maps are computed using the last feature maps of a segmentation network optimized only with global image-level labels. The proposed method can generate attention maps at full input resolution without need for interpolation during preprocessing, which allows small lesions to appear in attention maps. For comparison, we modify state-of-the-art methods to compute attention maps for weakly supervised object detection, by using a global regression objective instead of the more conventional classification objective. This regression objective optimizes the number of occurrences of the target object in an image, e.g. the number of brain lesions in a scan, or the number of digits in an image. We study the behavior of the proposed method in MNIST-based detection datasets, and evaluate it for the challenging detection of enlarged perivascular spaces – a type of brain lesion – in a dataset of 2202 3D scans with point-wise annotations in the center of all lesions in four brain regions. In MNIST-based datasets, the proposed method outperforms the other methods. In the brain dataset, the weakly supervised detection methods come close to the human intrarater agreement in each region. The proposed method reaches the best area under the curve in two out of four regions, and has the lowest number of false positive detections in all regions, while its average sensitivity over all regions is similar to that of the other best methods. The proposed method can facilitate epidemiological and clinical studies of enlarged perivascular spaces and help advance research in the etiology of enlarged perivascular spaces and in their relationship with cerebrovascular diseases.

Subject

Brain
Count
Deep learning
Detection
Enlarged perivascular spaces
Lesion
MRI
Perivascular spaces
Regression
Weak-labels
Weakly-supervised