Automatic Kidney Lesion Detection for CT Images Using Morphological Cascade Convolutional Neural Networks

Abstract

The CT scan image is one of the most useful tools for diagnosing and locating lesions in the kidney. It can provide precise information about the location and size of lesions in many medical applications. Manual and traditional medical testings are labor-consuming and time-costing. Nowadays, detecting lesions in CT automatically is an integral assignment to the paramount importance of clinical diagnosis. Computer-aided diagnosis (CAD) is needed to develop and improve medical testing efficiency. However, it is still a tremendous challenge to the extant low precision and incomplete detection algorithm. In this paper, we proposed a lesion detection tool using multi intersection over union (IOU) threshold based on morphological cascade convolutional neural networks (CNNs). For improving the detection of small lesions (1-5 mm) and increasing the stableness of network, we proposed two morphology convolution layers and modified feature pyramid networks (FPNs) in the faster RCNN and combined four IOU threshold cascade RCNNs. In this lesion detection task, the modified CNN was trained in pytorch framework. The experiments were conducted in CT kidney images of DeepLesion that are published by hospitals' picture archiving and communication systems (PACSs). Finally, our method achieved AP of 0.840 and AUC of 0.871, and the results demonstrated that our proposed detector is an outstanding tool for detecting lesions in CT and outperformed in the data set.INDEX TERMS Kidney lesion detect, deep learning, morphology, RCNN.

Vision transformer and explainable transfer learning models for auto detection of kidney cyst, stone and tumor from CT-radiography

“…The authors used 1,799 images there in total to train and validate the model. The authors in 30 proposed two morphology convolution layers, modified feature pyramid networks (FPNs) in the faster RCNN and combined four thresholds. They got an area under the curve (AUC) value of 0.871.…”

Neural Network-derived perfusion maps: a Model-free approach to computed tomography perfusion in patients with acute ischemic stroke

“…Considering the apparently similar performance of CNNs and deconvolution-based algorithms, one might ask why the former approach might be preferable. Although machine learning algorithms have largely proven to overcome conventional image processing algorithms in practically every field, applications to CTP imaging are still limited (segmentation,[20 22] noise reduction,[22 24] novelty detection,[23,25] radiation dose reduction[23]). In particular, up until now generation of synthetic maps has been done only with MRI DSC perfusion by Ho et al and Meier et al[26,27] Meier et al obtained results similar to ours: they compared the performance of a commercial FDA-approved perfusion software and a CNN not only to generate T-Max MRI perfusion maps, but also to identify selection criteria for reperfusion therapies.…”

“…Considering the apparently similar performance of CNNs and deconvolution-based algorithms, one might ask why the former approach might be preferable. Although machine learning algorithms have largely proven to overcome conventional image processing algorithms in practically every field, applications to CTP imaging are still limited (segmentation, [20][21][22] noise reduction, [22][23][24] novelty detection, [23,25] radiation dose reduction [23]). In particular, up until now generation of synthetic maps In this preliminary work, pre-processed images were used as input.…”

COVID-19: Role of Deep Learning and Cloud Through Identification of Kidney, Pancreas and Intestine

“…But there is no work in comparing the DL algorithms which achieves more accuracy in CKD. The major issues of COVID-19 which includes geographical issues, high-risk patient and recognition have been reviewed [8]. For detecting and finding location of lesions in the kidney, the CT scan image plays a major role as it provides useful information in many applications.…”