Deeply-Supervised 3D Convolutional Neural Networks for Automated Ovary and Follicle Detection from Ultrasound Volumes

Božidar Potočnik,Martin Šavc

doi:10.3390/app12031246

Abstract

Automated detection of ovarian follicles in ultrasound images is much appreciated when its effectiveness is comparable with the experts’ annotations. Today’s best methods estimate follicles notably worse than the experts. This paper describes the development of two-stage deeply-supervised 3D Convolutional Neural Networks (CNN) based on the established U-Net. Either the entire U-Net or specific parts of the U-Net decoder were replicated in order to integrate the prior knowledge into the detection. Methods were trained end-to-end by follicle detection, while transfer learning was employed for ovary detection. The USOVA3D database of annotated ultrasound volumes, with its verification protocol, was used to verify the effectiveness. In follicle detection, the proposed methods estimate follicles up to 2.9% more accurately than the compared methods. With our two-stage CNNs trained by transfer learning, the effectiveness of ovary detection surpasses the up-to-date automated detection methods by about 7.6%. The obtained results demonstrated that our methods estimate follicles only slightly worse than the experts, while the ovaries are detected almost as accurately as by the experts. Statistical analysis of 50 repetitions of CNN model training proved that the training is stable, and that the effectiveness improvements are not only due to random initialisation. Our deeply-supervised 3D CNNs can be adapted easily to other problem domains.

Highlights

A sexually mature female has two almond-shaped ovaries about the size of a large grape, one on each side of the uterus
This paper describes the development of two-stage deeply-supervised 3D Convolutional Neural Networks (CNN) based on the established U-Net
We have evaluated the effectiveness of selected follicle detection methods further using Sensitivity or Recall (S), Precision (P), Dice Similarity Coefficient (DSC), Jaccard Index (JCI) and the F1 score (F1), which are established metrics, but each of them covers only one aspect of the algorithm’s detection performance (On the other hand, our used evaluation protocol combines several aspects over several raters into a common assessment or overall score, respectively!)

Summary

Introduction

A sexually mature female has two almond-shaped ovaries about the size of a large grape, one on each side of the uterus. The human ovary consists of a surface, an inner medulla and outer cortex, with indistinct boundaries between the latter two. The medulla contains the blood vessels, lymphatic vessels and nerves, while the cortex embraces the developing follicles [1]. The follicle is certainly a very important part of the ovary. A few selected follicles begin to develop (grow) during each woman’s menstrual cycle. At the end of the menstrual cycle, typically, only one of these follicles reaches maturity and the rest deteriorate. This mature, so-called dominant, follicle breaks open and releases the egg from the ovary for possible fertilisation [1,2]

Objectives

Methods

Results

Discussion

Conclusion