An Automated Diagnosis Method for Lung Cancer Target Detection and Subtype Classification-Based CT Scans
Abstract
:1. Introduction
- (1)
- The improved method integrates a large separable kernel attention mechanism after the C2f module of the network. This strategy aims to enhance the model’s focus on lung cancer features. By introducing this mechanism into the Backbone section of the C2f layer, the model can more effectively identify and focus on key features in the input images while suppressing the influence of irrelevant features. Applying this attention mechanism in the Neck section of the C2f layer promotes information exchange among features of different scales, further improving the accuracy and robustness of feature representation.
- (2)
- The improved method combines depth-wise convolution with coordinate attention in the SPPF module, aiming to enlarge the model’s receptive field and reduce feature information loss caused by pooling operations. This improvement not only enhances the model’s focus on deep lung cancer features but also improves its perception of lung cancers with different shapes. Consequently, the model can more accurately detect and recognize various subtypes of lung cancer.
- (3)
- The improved method adopts MPDIOU Loss to replace CIOU Loss. To further enhance the accuracy of lung cancer detection and recognition, we introduce MPDIOU Loss to replace the original CIOU Loss. MPDIOU Loss can more accurately measure the differences between predicted boxes and ground truth boxes, thus optimizing the model’s target localization capabilities. By using this new loss function, the model can more accurately calculate IOU loss during the training process and adjust the model parameters accordingly, thereby achieving more precise and efficient detection and recognition of lung cancer subtypes.
2. Related Works
2.1. The Lung Cancer Detection Method Based on Manually Set Features
2.2. Lung Cancer Detection Methods Based on Traditional Machine Learning
2.3. The Lung Cancer Detection Method Based on Deep Learning
3. Method
3.1. C2f-LSKA Module
3.2. SPPF-CA Module
3.3. MPDIOU Loss
- (1)
- Overlap or Non-Overlap Region: Considers the overlap or non-overlap region between two bounding boxes to ensure the loss function responds appropriately to different scenarios.
- (2)
- Center Point Distance: Considers the distance between the center points of predicted and ground truth bounding boxes. By focusing on the position of the center points, MPDIOU can more accurately measure the relative positional relationship between the two bounding boxes.
- (3)
- Width and Height Deviation: Considers the deviation between the width and height of predicted bounding boxes and those of ground truth bounding boxes. This is necessary because, even if the aspect ratios are the same, the actual width and height values may differ, requiring correction for this deviation.
3.4. Evaluation Metrics
- (1)
- Precision: Precision, also known as positive predictive value, refers to the proportion of instances predicted as positive by the model that are actually positive. A high precision indicates that the model’s predictions of the positive class are more accurate. Precision is defined by Formula (27):
- (2)
- Recall: Recall, also known as sensitivity or true positive rate, is defined as the proportion of true positive instances to all actual positive instances. A high recall indicates that the model can detect positive samples comprehensively. Recall is defined by Formula (28):
- (3)
- Accuracy: Accuracy is defined as the proportion of correctly predicted samples to the total number of samples. A high accuracy indicates that the model can classify samples more accurately. Accuracy is defined by Formula (29):
- (4)
- Specificity: Specificity, also known as sensitivity or true negative rate, is typically refers to the identification accuracy of a test or diagnostic method for a specific disease or biological marker. Accuracy is defined by Formula (30):
4. Experimentation
4.1. Data Preprocessing
4.2. Sensitivity Analysis
4.3. Detection Comparative Experiment
4.4. Accuracy Comparison Experiment
4.5. Experimental Results Visualization
5. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Data Availability Statement
Conflicts of Interest
References
- Popper, H. Pathologic diagnosis of lung cancer–recent developments. Curr. Opin. Oncol. 2024, 36, 57–62. [Google Scholar] [CrossRef] [PubMed]
- Jin, M.; Tang, J.; Qin, J.; Huang, Y.; Wei, Y.; Zheng, J. An Epidemiological Analysis of Lung Cancer between 2002 and 2020. Acta Med. Sin. 2021, 34, 34–38. [Google Scholar]
- Del, A.; Franchi, P.; Contegiacomo, A.; Cicchetti, G.; Bonomo, L.; Larici, A. Missed lung cancer: When, where, and why? Diagn. Interv. Radiol. 2017, 23, 118. [Google Scholar]
- Hendrix, W.; Hendrix, N.; Scholten, E.T.; Mourits, M.; Trap-de Jong, J.; Schalekamp, S.; Korst, M.; van Leuken, M.; van Ginneken, B.; Prokop, M.; et al. Deep learning for the detection of benign and malignant pulmonary nodules in non-screening chest CT scans. Commun. Med. 2023, 3, 156. [Google Scholar] [CrossRef] [PubMed]
- Zhou, J.; Yu, B.; Guo, P.; Wang, S. The insufficiency of CT examination in early detection of central lung squamous cell carcinoma and squamous epithelial precancerous lesions. BMC Cancer 2024, 24, 299. [Google Scholar] [CrossRef] [PubMed]
- Rahal, H.R.; Slatnia, S.; Kazar, O.; Barka, E.; Harous, S. Blockchain-based multi-diagnosis deep learning application for various diseases classification. Int. J. Inf. Secur. 2024, 23, 15–30. [Google Scholar] [CrossRef]
- Messay, T.; Hardie, R.; Rogers, S. A new computationally efficient CAD system for pulmonary nodule detection in CT imagery. Med. Image Anal. 2010, 14, 390–406. [Google Scholar] [CrossRef]
- Paing, M.P.; Choomchuay, S. A computer aided diagnosis system for detection of lung nodules from series of CT slices. In Proceedings of the 2017 14th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology (ECTI-CON), Phuket, Thailand, 27–30 June 2017; IEEE: Piscataway, NJ, USA, 2017; pp. 302–305. [Google Scholar]
- Gupta, A.; Martens, O.; Le Moullec, Y.; Saar, T. Methods for increased sensitivity and scope in automatic segmentation and detection of lung nodules in CT images. In Proceedings of the 2015 IEEE International Symposium on Signal Processing and Information Technology (ISSPIT), Abu Dhabi, United Arab Emirates, 7–10 December 2015; IEEE: Piscataway, NJ, USA, 2015; pp. 375–380. [Google Scholar]
- Soni, A.; Rai, A.; Shivhare, E. Atomatic Lung Cancer Detection using Sobel & Morphological Operations. Int. J. Eng. Adv. Technol. 2019, 8, 2772–2777. [Google Scholar]
- Sairam, P. Lung Cancer Edge Detection Using Canny Edge Detector. Int. J. Ind. Electron. Control. 2014, 6, 121–126. [Google Scholar]
- Rezaie, A.; Habiboghli, A. Detection of Lung Nodules on Medical Images by the Use of Fractal Segmentation. Int. J. Interact. Multimed. Artif. Intell. 2017, 4, 15–19. [Google Scholar]
- Lian, L. Intelligent Optimization Algorithm for Support Vector Machine: Research and Analysis of Prediction Ability. Int. J. Artif. Intell. Tools 2024, 33, 2350048. [Google Scholar] [CrossRef]
- Leo, B. Random forests. Mach. Learn. 2001, 45, 5–32. [Google Scholar]
- Johora, F.T.; Jony, M.H.; Hossain, M.S.; Rana, H.K. Lung Cancer Detection Using Marker-Controlled Watershed with Svm. GUB J. Sci. Eng. (GUBJSE) 2018, 5, 14–30. [Google Scholar] [CrossRef]
- Firdaus, Q.; Sigit, R.; Harsono, T.; Anwar, A. Lung cancer detection based on CT-scan images with detection features using gray level co-occurrence matrix (GLCM) and support vector machine (SVM) methods. In Proceedings of the 2020 International Electronics Symposium (IES), Los Angeles, CA, USA, 21–26 July 2020; IEEE: Piscataway, NJ, USA, 2020; pp. 643–648. [Google Scholar]
- Khordehchi, E.A.; Ayatollahi, A.; Daliri, M.R. Automatic lung nodule detection based on statistical region merging and support vector machines. Image Anal. Stereol. 2017, 36, 65–78. [Google Scholar] [CrossRef]
- Redmon, J.; Divvala, S.; Girshick, R.; Farhadi, A. You only look once: Unified, real-time object detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27–30 June 2016; pp. 779–788. [Google Scholar]
- Redmon, J.; Farhadi, A. YOLO9000: Better, faster, stronger. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, HI, USA, 21–26 July 2017; pp. 7263–7271. [Google Scholar]
- Redmon, J.; Farhadi, A. Yolov3: An incremental improvement. arXiv 2018, arXiv:1804.02767. [Google Scholar]
- Liu, W.; Anguelov, D.; Erhan, D.; Szegedy, C.; Reed, S.; Fu, C.Y.; Berg, A.C. Ssd: Single shot multibox detector. In Proceedings of the Computer Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, 11–14 October 2016; Proceedings, Part I 14. Springer International Publishing: Berlin/Heidelberg, Germany, 2016; pp. 21–37. [Google Scholar]
- Lin, Q.; Chen, X.; Liu, L.; Cao, Y.; Man, Z.; Zeng, X.; Huang, X. Detecting multiple lesions of lung cancer-caused metastasis with bone scans using a self-defined object detection model based on SSD framework. Phys. Med. Biol. 2022, 67, 225009. [Google Scholar] [CrossRef] [PubMed]
- Loraksa, C.; Mongkolsomlit, S.; Nimsuk, N.; Uscharapong, M.; Kiatisevi, P. Development of the osteosarcoma lung nodules detection model based on SSD-VGG16 and competency comparing with traditional method. IEEE Access 2022, 10, 65496–65506. [Google Scholar] [CrossRef]
- Mammeri, S.; Amroune, M.; Haouam, M.-Y.; Bendib, I.; Silva, A.C. Early detection and diagnosis of lung cancer using YOLO v7, and transfer learning. Multimed. Tools Appl. 2023, 83, 30965–30980. [Google Scholar] [CrossRef]
- Ji, Z.; Zhao, J.; Liu, J.; Zeng, X.; Zhang, H.; Zhang, X.; Ganchev, I. ELCT-YOLO: An efficient one-stage model for automatic lung tumor detection based on CT images. Mathematics 2023, 11, 2344. [Google Scholar] [CrossRef]
- Ren, S.; He, K.; Girshick, R.; Sun, J. Faster R-CNN: Towards real-time object detection with region proposal networks. IEEE Trans. Pattern Anal. Mach. Intell. 2016, 39, 1137–1149. [Google Scholar] [CrossRef]
- He, K.; Gkioxari, G.; Dollár, P.; Girshick, R. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision, Venice, Italy, 2–29 October 2017; pp. 2961–2969. [Google Scholar]
- Su, Y.; Li, D.; Chen, X. Lung nodule detection based on faster R-CNN framework. Comput. Methods Programs Biomed. 2021, 200, 105866. [Google Scholar] [CrossRef] [PubMed]
- Nguyen, C.C.; Tran, G.S.; Nguyen, V.T.; Burie, J.-C.; Nghiem, T.P. Pulmonary nodule detection based on faster R-CNN with adaptive anchor box. IEEE Access 2021, 9, 154740–154751. [Google Scholar] [CrossRef]
- Xu, J.; Ren, H.; Cai, S.; Zhang, X. An improved faster R-CNN algorithm for assisted detection of lung nodules. Comput. Biol. Med. 2023, 153, 106470. [Google Scholar] [CrossRef] [PubMed]
- Saad, M.; Choi, T.S. Deciphering unclassified tumors of non-small-cell lung cancer through radiomics. Comput. Biol. Med. 2017, 91, 222–230. [Google Scholar] [CrossRef] [PubMed]
- Liu, J.; Cui, J.; Liu, F.; Yuan, Y.; Guo, F.; Zhang, G. Multi-subtype classification model for non-small cell lung cancer based on radiomics: SLS model. Med. Phys. 2019, 46, 3091–3100. [Google Scholar] [CrossRef] [PubMed]
- Wang, S.; Dong, L.; Wang, X.; Wang, X. Classification of pathological types of lung cancer from CT images by deep residual neural networks with transfer learning strategy. Open Med. 2020, 15, 190–197. [Google Scholar] [CrossRef] [PubMed]
- Pang, S.; Zhang, Y.; Ding, M.; Wang, X.; Xie, X. A deep model for lung cancer type identification by densely connected convolutional networks and adaptive boosting. IEEE Access 2019, 8, 4799–4805. [Google Scholar] [CrossRef]
- Han, Y.; Ma, Y.; Wu, Z.; Zhang, F.; Zheng, D.; Liu, X.; Tao, L.; Liang, Z.; Yang, Z.; Li, X.; et al. Histologic subtype classification of non-small cell lung cancer using PET/CT images. Eur. J. Nucl. Med. Mol. Imaging 2021, 48, 350–360. [Google Scholar] [CrossRef] [PubMed]
- Gao, Y.; Song, F.; Zhang, P.; Liu, J.; Cui, J.; Ma, Y.; Zhang, G.; Luo, J. Improving the subtype classification of non-small cell lung cancer by elastic deformation based machine learning. J. Digit. Imaging 2021, 34, 605–617. [Google Scholar] [CrossRef]
- Jacob, C.; Menon, G. Pathological categorization of lung carcinoma from multimodality images using convolutional neural networks. Int. J. Imaging Syst. Technol. 2022, 32, 1681–1695. [Google Scholar] [CrossRef]
- Marentakis, P.; Karaiskos, P.; Kouloulias, V.; Kelekis, N.; Argentos, S.; Oikonomopoulos, N.; Loukas, C. Lung cancer histology classification from CT images based on radiomics and deep learning models. Med. Biol. Eng. Comput. 2021, 59, 215–226. [Google Scholar] [CrossRef] [PubMed]
- Kosaraju, S.; Park, J.; Lee, H.; Yang, J.W.; Kang, M. Deep learning-based framework for slide-based histopathological image analysis. Sci. Rep. 2022, 12, 19075. [Google Scholar] [CrossRef] [PubMed]
- Barbouchi, K.; El Hamdi, D.; Elouedi, I.; Aïcha, T.B.; Echi, A.K.; Slim, I. A transformer-based deep neural network for detection and classification of lung cancer via PET/CT images. Int. J. Imaging Syst. Technol. 2023, 33, 1383–1395. [Google Scholar] [CrossRef]
Algorithm YOLO V8-LCM |
Input: input CT images and their annotations. |
Output: Model weights and detection indicators. |
for ith-fold, (train-dataset, val-dataset) in K-fold dataset: |
for jth-epoch in range(epochs): |
for step, train-data in enumerate (train-dataset): |
Using train-data to train weights using YOLO V8-LCM model in Section 3. |
Calculate the classification loss and MPDIOU Loss and return the gradient. |
Update parameters in the network in this step. |
end |
Save the weights of the optimal model in this jth-epoch. |
end |
Verify on the val-dataset and calculate detection indicators in this ith-fold. |
end |
Calculate the mean and standard deviation of the detection indicators in the K-fold dataset. |
Classess | Precision | Recall | mAP (50%) | mAP (50–90%) |
---|---|---|---|---|
ALL-LESIONS | 0.954 ± 0.0028 | 0.968 ± 0.0037 | 0.981 ± 0.0052 | 0.629 ± 0.0152 |
Class 0 (SCLC) | 0.963 ± 0.0023 | 0.958 ± 0.0029 | 0.986 ± 0.0037 | 0.619 ± 0.0131 |
Class 1 (LCC) | 0.961 ± 0.0025 | 0.947 ± 0.0032 | 0.981 ± 0.0028 | 0.592 ± 0.0093 |
Class 2 (ADC) | 0.952 ± 0.0031 | 0.951 ± 0.0026 | 0.966 ± 0.0029 | 0.658 ± 0.0142 |
Class 3 (SCC) | 0.967 ± 0.0027 | 0.979 ± 0.0028 | 0.986 ± 0.0033 | 0.663 ± 0.0152 |
Module | Precision | Recall | Specificity | mAP (50%) | mAP (50–95%) |
---|---|---|---|---|---|
YOLO V8 | 0.942 ± 0.0047 | 0.953 ± 0.0073 | 0.958 ± 0.0089 | 0.968 ± 0.0061 | 0.589 ± 0.0173 |
YOLO V8-CM | 0.945 ± 0.0036 | 0.956 ± 0.0057 | 0.962 ± 0.0062 | 0.972 ± 0.0058 | 0.617 ± 0.0147 |
YOLO V8-LM | 0.949 ± 0.0042 | 0.967 ± 0.0053 | 0.96 ± 0.0059 | 0.976 ± 0.0051 | 0.619 ± 0.0136 |
YOLO V8-LC | 0.951 ± 0.0031 | 0.963 ± 0.0066 | 0.961 ± 0.0057 | 0.97 ± 0.0055 | 0.6138 ± 0.0143 |
YOLO V8-LCM | 0.954 ± 0.0028 | 0.968 ± 0.0052 | 0.969 ± 0.0061 | 0.981 ± 0.0052 | 0.629 ± 0.0152 |
Model | Precision | Recall | Specificity | mAP (50%) | Parameters | Speed |
---|---|---|---|---|---|---|
SSD | 0.896 ± 0.013 | 0.933 ± 0.013 | 0.922 ± 0.019 | 0.939 ± 0.0168 | 134 M | 5.3 ms |
Faster R-CNN | 0.907 ± 0.018 | 0.926 ± 0.018 | 0.923 ± 0.021 | 0.957 ± 0.013 | 430 M | 12.2 ms |
YOLO V3 | 0.918 ± 0.0087 | 0.958 ± 0.012 | 0.955 ± 0.083 | 0.965 ± 0.0098 | 32 M | 4.5 ms |
YOLO V5 | 0.925 ± 0.0066 | 0.956 ± 0.0066 | 0.961 ± 0.0068 | 0.973 ± 0.0068 | 1.8 M | 2.9 ms |
YOLO V7 | 0.919 ± 0.0067 | 0.947 ± 0.0078 | 0.956 ± 0.0078 | 0.965 ± 0.0072 | 6.03 M | 3.8 ms |
YOLO V8 | 0.922 ± 0.0051 | 0.951 ± 0.0069 | 0.955 ± 0.0086 | 0.969 ± 0.0077 | 3.1 M | 3.3 ms |
YOLO V8-LCM | 0.934 ± 0.0035 | 0.958 ± 0.0058 | 0.966 ± 0.0065 | 0.977 ± 0.0065 | 3.2 M | 3.6 ms |
References | Data Type | Database | Subtypes of Lung Cancer | Samples | Algorithm Model | Results |
---|---|---|---|---|---|---|
Saad et al. [31] | CT | Lung1 | ADC, SCC, LCC | 317 | SVM | ACC = 0.78 |
Liu et al. [32] | CT | Lung1 | ADC, SCC, LCC, NOS | 349 | SVM | ACC = 0.86 |
Wang et al. [33] | CT | Private dataset | SCC, SQC, IA, ISA | 168 | ResNet | ACC = 0.86 |
Pang et al. [34] | CT | Private dataset | ADC, SCC, SQC | 1183 | DenseNet | ACC = 0.90 |
Han et al. [35] | CT | Private dataset | ADC, SQC | 283 | VGG16 | ACC = 0.84 |
Gao et al. [36] | CT | Lung1 | SCC, LCC | 169 | GBRT | ACC = 0.96 |
Jacob et al. [37] | PET/CT | Lung-PET-CT-Dx | ADC, SCC, SQC | 204 | Shadow CNN | ACC = 0.95 |
Wang et al. [32] | PET/CT | Lung-PET-CT-Dx | ADC, SCC, SQC | 232 | ResNet | ACC = 0.81 |
Marentakis et al. [38] | CT | TCIA | ADC, SCC | 914 | LSTM+ Inception | AUC = 0.74 |
Kosaraju et al. [39] | WSI | GNUH | SCC, ADC, SCLC, LCNEC | 40,000 patches | CAT-Net | Micro F1 = 0.701 |
Khalil et al. [40] | PET/CT | Lung-PET-CT-Dx | ADC, SCC, SQC | 270 | MobiNet | ACC = 0.74 |
Khalil et al. [40] | PET/CT | Lung-PET-CT-Dx | ADC, SCC, SQC | 270 | Xception | ACC = 0.78 |
Khalil et al. [40] | PET/CT | Lung-PET-CT-Dx | ADC, SCC, SQC | 270 | DETR | ACC = 0.96 |
Our approach | CT | Lung-PET-CT-Dx, NSCLC-Radiomics, NSCLC-Radiogenomics | ADC, SQC, SCC, LCC | 814 | YOLO V8-LCM | ACC = 0.961 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Wang, L.; Zhang, C.; Zhang, Y.; Li, J. An Automated Diagnosis Method for Lung Cancer Target Detection and Subtype Classification-Based CT Scans. Bioengineering 2024, 11, 767. https://doi.org/10.3390/bioengineering11080767
Wang L, Zhang C, Zhang Y, Li J. An Automated Diagnosis Method for Lung Cancer Target Detection and Subtype Classification-Based CT Scans. Bioengineering. 2024; 11(8):767. https://doi.org/10.3390/bioengineering11080767
Chicago/Turabian StyleWang, Lingfei, Chenghao Zhang, Yu Zhang, and Jin Li. 2024. "An Automated Diagnosis Method for Lung Cancer Target Detection and Subtype Classification-Based CT Scans" Bioengineering 11, no. 8: 767. https://doi.org/10.3390/bioengineering11080767