INTERVERTEBRAL DISC DETECTION IN X-RAY IMAGES USING FASTER R-CNN : A DEEP LEARNING APPROACH

Automatic identification of specific osseous landmarks on the spinal radiograph can be used to automate calculations for diagnosing ligament instability and injury, which affect 75% of patients injured in motor vehicle accidents and is a precursor for other related diseases. In this work, we propose to use deep learning based object detection method as the first step towards identifying landmark points. The significant breakthrough of deep learning technology has made it a prevailing choice for perception based applications, however, the lack of large annotated training dataset has brought challenges to utilize the technology in medical image processing field. In this work, we try to address this problem by fine-tuning a deep network, in particular Faster-RCNN, a state-of-the-art deep detection network in natural image domain, using small annotated clinical datasets. In the experiment we show that, by using only 81 lateral lumbar X-Ray training images, as shown in Fig. 1, one can achieve satisfactory results. We achieved high average precision of 0.651, evaluated by widely used VOC2007 evaluation metric and the time is reduced to 0.2 second per image, which is significantly faster than traditional sliding window based classification techniques.

脊柱X光片上特定骨骼标志的自动识别可用于自动计算以诊断韧带不稳定和损伤，这些情况影响75%的机动车事故受伤患者，并且是其他相关疾病的先兆。在这项工作中，我们提议使用基于深度学习的目标检测方法作为识别标志点的第一步。深度学习技术的重大突破使其成为基于感知的应用的主流选择，然而，缺乏大量带注释的训练数据集给在医学图像处理领域应用该技术带来了挑战。在这项工作中，我们试图通过微调一个深度网络来解决这个问题，特别是使用小型带注释的临床数据集对自然图像领域中最先进的深度检测网络Faster - RCNN进行微调。在实验中我们表明，仅使用81张腰椎侧位X光训练图像（如图1所示），就可以取得令人满意的结果。通过广泛使用的VOC2007评估指标评估，我们获得了0.651的高平均精度，并且每张图像的处理时间减少到0.2秒，这比传统的基于滑动窗口的分类技术快得多。