Original InvestigationCT Image-based Decision Support System for Categorization of Liver Metastases Into Primary Cancer Sites: Initial Results
Introduction
Computed tomography (CT) is one of the most common modalities used for the detection, diagnosis, and follow-up of liver lesions (1), with the images acquired before and after intravenous injection of a contrast agent (Iohexol 350 mg/mL up to 2 mL/kg) and oral contrast agent (Iohexol 350 mg/mL 52 mL diluted in 2 L water). Radiologists usually detect and diagnose liver lesions based on the different density of the lesions at different time points in the scan.
The liver is one of the most common organs to develop metastases. Metastatic liver lesions can be derived from different primary cancer types, including melanoma, breast cancer, colon cancer, neuroendocrine, pancreatic cancer, and others. Not infrequently when liver metastases are detected, the radiologists do not know the location of the primary site of cancer. In these cases, a broad search for the source of the primary tumor is performed. In some cases, the source of the cancer remains unknown, despite a thorough physical examination including breast, pelvic and rectal examination, basic laboratory tests, urinalysis, X-rays, CT scans, and other imaging studies. Knowing the source of the primary tumor is important to help localize the search, for example, colonoscopy for colon carcinoma, gastroscopy for gastric carcinoma, etc, and to treat it accordingly.
Finding the primary site can be a time-consuming task requiring multiple examinations. The difficulty of this task highlights the need for assistive computerized tools in this domain. Several studies that categorize metastatic tumours by taking biopsies and using molecular gene expression profiling to predict the tissue of origin can be found 2, 3, 4. Segal et al. (5) showed that dynamic imaging traits in CT examinations systematically correlate with the global gene expression programs of primary human liver cancer. To the best of our knowledge, no methods to identify the source of the metastasis based on its appearance, directly from the CT image scan, currently exist.
The current work focuses on such a solution. We present an automated algorithm for categorization of liver metastases to their primary cancer sites using CT examinations following existing radiological workflow procedures. The method proposed is presented in the Methods section. Experimental results are shown in the Experiments and Results section. The experiments include the derivation of the final classification framework used in this work. In the Discussion section, a discussion of the results is presented followed by a conclusion of the work.
Section snippets
Data Description
The data used in the current research include CT scans from a large Medical Center1 from 2009 to 2014. Different CT scanners were used with 0.7090- to 1.1719-mm pixel spacing, 1.25- to 5-mm slice thickness, 120 kVp, and different convolution kernels (GE—standard/CHST, Siemens—B31s, Phillips—B/C). The scans were selected and marked by a radiologist. They include 100 patients with 192 lesions. A development set was used to establish the proposed
Experiments and Results
We evaluated the class separability using the top-n accuracy in classification using LOOCV. Here, for each patient, the top n guesses (ie, the classes with the top n probabilities) are considered. The number of times a correct classification label appears in the top n guesses is divided by the total number of test samples. In clinical practice, the classification results can suggest the order of examinations; hence, in our case, the top-1, top-2, and top-3 accuracy results are all of interest.
Discussion
The descriptors used in this study are not always easy to see by the naked eye, especially when dealing with NC CT examinations. Metastases from different primary sites can look very much alike. However, our results show that texture characteristics can help distinguish between metastases that derived from different primary sites. From Table 2, we conclude that the texture features are very informative for the categorization task. Another interesting observation is that CNN-based features were
Conclusions
Nowadays, the primary site of a metastatic lesion is usually determined by performing multiple examinations until the primary cancer is found. Alternative works have shown that the primary site can be determined using a biopsy. Our objective in the current work was to explore the possibility of using portal phase and NC CT scans and to provide a noninvasive automatic methodology for the categorization of liver metastatic lesions to their primary cancer site. This system can help prioritize the
Acknowledgment
This research was supported by the Israel Science Foundation (grant No. 1918/16).
Avi Ben-Cohen's scholarship was funded by the Buchmann Scholarships Fund.
References (22)
- et al.
Blinded comparator study of immunohistochemical analysis versus a 92-gene cancer classifier in the diagnosis of the primary site in metastatic tumors
J Mol Diagn
(2013) - et al.
A comparative study of texture measures with classification based on featured distributions
Pattern Recognit
(1996) - et al.
Body CT and oncologic imaging 1
Radiology
(2000) - et al.
Molecular gene expression profiling to predict the tissue of origin and direct site-specific therapy in patients with carcinoma of unknown primary site: a prospective trial of the Sarah Cannon Research Institute
J Clin Oncol
(2013) - et al.
A multicenter study directly comparing the diagnostic accuracy of gene expression profiling and immunohistochemistry for primary site identification in metastatic tumors
Am J Surg Pathol
(2013) - et al.
Decoding global gene expression programs in liver cancer by noninvasive imaging
Nat Biotechnol
(2007) - et al.
Automated retrieval of CT images of liver lesions on the basis of image similarity: method and preliminary results 1
Radiology
(2010) - et al.
Assessing the classification of liver focal lesions by using multi-phase computer tomography scans
- et al.
Computer-aided diagnosis of liver tumors based on multi-image texture analysis of contrast-enhanced CT selection of the most appropriate texture features
Stud Log Grammar Rhetor
(2013) - et al.
Unsupervised texture segmentation using Gabor filters
Measures of acutance and shape for classification of breast tumors
IEEE Trans Med Imaging
Cited by (30)
Artificial intelligence
2023, Translational Sports MedicineThe application of artificial intelligence in hepatology: A systematic review
2022, Digestive and Liver DiseaseCitation Excerpt :A total of 150 studies were selected. However, the object of this systematic review will be the 66 articles describing the application of AI approaches developed to improve hepatocellular carcinoma (HCC) and liver metastasis diagnostics (Supplementary Table 1) [5-70], and the 41 articles that use AI to improve non-alcoholic fatty liver disease (NAFLD) and fibrosis diagnostics (Table 1 and Supplementary Table 2) [71-111]. Since the wide diversity of objectives, methods, and metrics precluded a quantitative approach, we performed a clustering of manuscripts based on the type of data that have been used to construct the algorithms.
Machine learning based liver disease diagnosis: A systematic review
2022, NeurocomputingCitation Excerpt :We have briefly discussed the best studies with respect to each modality and the number of classes in the following paragraphs. Studies [227–230] have opted variant methodologies to accomplish clinical diagnostic tasks. After preprocessing, liver disease diagnosis requires unique features to classify various diseases.
Artificial Intelligence in Liver Diseases: Recent Advances
2024, Advances in TherapyReview on enhancing clinical decision support system using machine learning
2024, CAAI Transactions on Intelligence Technology
Avi Ben-Cohen and Eyal Kang contributed equally.
Michal Marianne Amitai and Hayit Greenspan contributed equally.