|
Ref. | Purposes | Challenges |
|
[36] | Deep learning techniques are used to detect, segment, and classify pulmonary nodules in CT scans | Generalization ability problem for learning-based methods. It happens because of the different training datasets and the methods. |
[29] | A comprehensive analysis of deep learning with convolutional neural network (CNN) methods and their performances | Problems with the generalizability and explication of the detection results, lack of accurate clinical decision-making tools, and well-labeled medical datasets |
[30] | The review of recent studies in lung nodule detection and classification provides an insight into technological advancements | Low sensitivity, high false positive rate, time-consuming, small database, poor performance rates, and so on |
[4] | A comparison of various machine learning-based methods for detecting lung cancer has been presented | Mainly focuses on machine learning techniques for classification rather than other processes. Also avoid the MRI type data. |
[31] | Review of recent deep learning algorithms and architectures for lung cancer detection | The data and the unbalanced nature of it are the current limitations |
[37] | Discussing the most recent developments in the field | The size of the target object within the image makes it difficult to implement a CNN; as the size of the target object varies, studies proposed training the model with images of varying scales to teach the model about this size variation |
[38] | Providing an accurate diagnosis and prognosis is essential in lung cancer treatment selection and planning | Incorporating knowledge from clinical and biological studies into deep learning methods and utilizing and integrating multiple medical imaging methods |
[27] | Algorithms used for each processing step are presented for some of the most current state-of-the-art CAD systems | Limitation of more interactive systems that allow for better use of automated methods in CT scan analysis |
[33] | An overview of the current state-of-the-art deep learning-aided lung cancer detection methods, as well as their key concepts and focus areas | Limited datasets and high correlation of errors in handling large image sizes |
[35] | A summary of existing CAD approaches for preprocessing, lung segmentation, false positive reduction, lung nodule detection, segmentation, classification, and retrieval using deep learning on CT scan data | Deficient data annotation, overfitting, lack of interpretability, and uncertainty quantification (UQ) |
[39] | A survey of what CADe schemes are used to detect pulmonary nodules will help radiologists make better diagnoses | Slight increase in lung density and micronodules whose diameters are less than 3 mm are difficult to detect. For multimodality, clinical records and medical images are not combined. |
|