CT texture analysis of perihilar cholangiocarcinoma – associations with tumor grading, tumor markers and clinical outcome

doi:10.21203/rs.3.rs-2982619/v1

Download PDF

Research Article

CT texture analysis of perihilar cholangiocarcinoma – associations with tumor grading, tumor markers and clinical outcome

https://doi.org/10.21203/rs.3.rs-2982619/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Background:

Texture analysis derived from computed tomography (CT) might be able to provide clinically relevant imaging biomarkers and might be associated with histopathology features in tumors. The present study sought to elucidate possible associations between texture features derived from CT images with grading, tumormarkers and survival in extrahepatic, perihilar cholangiocarcinomas tumors.

Methods:

22 patients (n = 10 females; 45%, mean age of 71.8 ± 8.7 years) were included into this retrospective study. Texture analysis was performed using the free available Mazda software. All tumors were histopathologically confirmed.

Results:

In discrimination analysis, “S(1,1)SumVarnc” was statistically significantly different between patients with long-term survival and non long-term survival (mean 275.80 ± 32.57 vs. 239.65 ± 26.03, p = 0.01). The first-order parameter “Skewness” was associated with the tumormarker „Carcinoembryonic antigen “(CEA) (r=-0.66, p = 0.01). A statistically significant correlation of the texture parameter “S(5,0)SumVarnc” with tumor grading was identified (r=-0.587, p = 0.004). Several other texture features were correlated with the tumormarkers “Ca-19-9” and “AFP” and with T- and N-stadium of the tumors.

Conclusion:

Several texture features derived from CT were associated with tumor characteristics and survival in patients with perihilar cholangiocarcinomas. CT texture features could be used as valuable novel biomarkers in clinical routine.

Computed tomography

texture analysis

Klatskin tumor

extrahepatic cholangiocarcinoma

As an emergent tool of imaging analysis, texture analysis allows the quantification of radiological images, which might aid as novel imaging biomarkers (1–3). Especially images derived from computed tomography (CT) are commonly used to perform this technique and are in particular suitable due to the robust imaging acquisition and high availability (1–3).

Texture analysis on one hand consists of the calculation of first order features, which are also known as histogram features. These are being calculated by issuing imaging features into histograms without concern of spatial relationships (1). These parameters, comprising skewness, kurtosis and entropy are able to reflect different histopathology parameters in several tumor entities (3–6). Besides histogram features, also second order statistics can be calculated, the so-called texture features (1). These are described by the spatial relationship between voxels with similar gray levels within a lesion. Therefore, texture features provide a measure of intralesional heterogeneity. As higher heterogeneity of tumors has been shown to correlate with poorer patient survival (7), texture features might prove even more helpful as an imaging biomarker when compared to histogram features.

Cholangiocarcinoma is a malignant disease occurring along side the biliary tree. Commonly, Cholangiocarcinomas are classified according to their anatomical subtype. Intrahepatic and extrahepatic cholangiocarcinomas, the latter further being divided in perihilar and distal cholangiocarcinomas (8). This tumor entitity has an overall poor prognosis and has frequent recurrence after surgical treatment. Relapse-free survival for extrahepatic cholangiocarcinomas is reported to be just up to 52% in the first 2 years and down to 38% after 5 years (9). The median survival is nowadays ranging from only 19 to 39 months (10).

Due to this very aggressive nature, the clinical management remains a challenge. Conventional staging consists of cross sectional imaging with CT and MRI imaging, typically carried out with contrast agent application and scanning in arterial and portal venous phase. The diagnostic sensitivity ranges from 75–92% (11,12). However, conventional CT images can only offer information regarding possible resectability and tumor spread (13).

CT texture analysis might provide further prognostic information helpful for a more detailed staging and therefore also further clinical information.

The principal hypothesis of this work is that texture analysis derived from CT-images might be able to correlate with biological characteristics of the perihilar cholangiocarcinomas comprising of tumor grading, TNM-stadium, tumormarker levels and postoperative patient survival. In that case, texture analysis could possibly represent a useful tool in clinical routine to further characterize these tumors non-invasively during the initial tumor staging and to even improve the clinical managment.

2.1. Study design

For this retrospective, observational study, all procedures performed involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. It received ethical approval from the local ethics committee at the medical faculty of the university of Leipzig (EK: 243-14-14072014). All participants provided written informed consent of this post-hoc scientific analysis and the publication.

The radiological database of the University hospital of Leipzig was retrospectively screened between 2016 and 2022 for availabe and suitable imaging data of perihilar extrahepatic Cholangiocarcinomas.

Inclusion criteria were sufficient presurgical CT data, laborchemical data and performed surgical resection with histopathological examination and follow-up survival data.

Overall, 22 patients (n = 10 females, 45%) with a mean age of 71.8 ± 8.8 years and histopathological proven perihilar cholangiocarcinoma were enrolled. 30-day mortality and long-term survival was evaluated as primary study outcomes.

2.2. Surgical treatment

All included patients underwent surgical treatment between 2017 and 2021 with curative intent. The used nomenclature of hepatic anatomy and resection was in compliance with the Brisbane 2000 classification system. A series of anatomic left and right hemihepatectomies was carried out with or without wedge resections of neighbouring liver segments. In some of the cases, extended left and right hemihepatectomies commonly referred to as trisectionectomy were performed. En-bloc resection of segment 1 was case dependent (14,15). Roux-en-Y anastomosis was used to reconstruct the gastrointestinal tract. Furthermore, regional lymphadenectomie was performed in the lymph node groups 1, 2 and 3 (16). Postoperative follow-ups of the treated patients were performed in a clinical outpatient- setting.

2.3. Histopathology and laboratory analyses

All resected perihilar cholangiocarcinomas were histopathologically analyzed blinded to the imaging results in clinical routine.

The tumormarkers Carcinoembryonic antigen (CEA) (available in n = 13 patients, 59.1%), Carbohydrate antigen 19 − 9 (CA 19 − 9) (available in n = 14 cases, 63.6%) and alpha-fetoprotein (AFP) (available in n = 8 of the cases, 36.4%) were determined before the surgical treatment in a clinical setting. Preoperative bilirubine and quick-values were also extracted.

2.4. Image analysis

CT imaging was performed in a clinical setting with a 256 slices CT scanner (iCT, Philips, Amsterdam, Netherlands) after intravenous application of 90 ml iodinated intravenous contrast medium (injected at a rate of 1.5 to 3.5 ml/s by a power injector (Medtron GmbH, Germany), with a scan delay of 70 seconds after onset of injection. The portalvenous phase images were evaluated for all patients. Typical imaging parameters were 120 kVp and 150 to 300 mAs. Reconstructed slice thickness was 1 mm.

2.5. Texture analysis

CT images were transformed into DICOM format. Further processing of the images was performed with the free available texture analysis software MaZda (version 4.7, available at http://www.eletel.p.lodz.pl/mazda/) (17,18). A polygonal ROI was placed on the representative axial slice with the largest tumor diameter. The ROI was drawn clearly within the boundary of the tumor. Images were analyzed in 1 mm abdomen kernel reconstructed slices. ROI placement was executed in a blinded manner to the clinical features. For each ROI, gray-level normalization was implemented, using the limitation of dynamics to µ ± 3 SD (µ gray level mean, SD standard deviation) to minimize the influence of contrast and brightness variation, as described in previous studies (6,19). The following features were extracted: gray-level histogram co-occurrence matrix (angular second moment, contrast, correlation, entropy, sum entropy, sum of squares, sum average, sum variance, inverse difference moment, difference entropy, difference variance (for four directions), run-length matrix (run-length non-uniformity, gray-level non-uniformity, long run emphasis, short run emphasis, fraction of image in runs)), absolute gradient, autoregressive model (theta 1 to 4, sigma), and wavelet transform. A total amount of 276 texture features were retrieved from every tumor. Figure 1 displays representative cases of the patient sample.

2.6. Statistical analysis

Statistical analysis was performed using SPSS Statistics 29 (IBM, Armonk, New York, USA). Collected data were evaluated by means of descriptive statistics (absolute and relative frequencies). Spearman's correlation coefficient (r) was used to analyze associations between texture features and grading, TNM and tumormarkers. Differences between survival and non-survival were investigated by two-tailed Mann–Whitney test. In all instances, p-values < 0.05 were taken to indicate statistical significance.

An overview of the investigated patient sample is provided by Table 1. Patients with long-time survival were being followed-up without signs of tumor recurrence for a mean time of 33.3 ± 20.3 months (ranging from 8 to 61 months).

Table 1

Overview of the descriptive statistics of the included patient sample.
	Survivors (n = 10)	Non-survivors (n = 12)	p-value
Gender (female, %)	5 (50)	5 (42)	0.77
Age, years	71.90 ± 7.40	71.67 ± 10.07	0.87
Bismuth classification			0.82
I	0	0
II	0	1
IIIA	1	0
IIIB	0	1
IV	9	10
Grading
I	0	1
II	8	9
III	2	2
pT			0.08
pT1a	0	1
pT1b	0	2
pT2a	2	5
pT2b	6	2
pT3	1	1
pT4	1	1
N			0.38
N0	5	8
N1	3	4
N2	2	0
M			0.77
M0	10	11
M1	0	1
Preoperative Bilirubine (µmol/l)	11.15 ± 3.18	34.58 ± 16.78	0.03
Preoperative Quick (%)	86.50 ± 20.51	86.83 ± 16.69	0.92
Preoperative CEA	1.91 ± 0.46	4.67 ± 5.99	0.83
Preoperative CA 19 − 9	111.50 ± 38.89	221.10 ± 163.44	0.66
Preoperative AFP	2.43 ± 0.86	10.29 ± 17.76	0.29

3.1 Discrimination analysis of survival

In discrimination analysis for clinical outcome, the following CT texture frautes were statistically different between 30-day mortality and survival: S(0,1)SumOfSqs (30-day-survival: 108.74 ± 3.88 vs. Non-survival: 102.99 ± 3.04, p = 0.035), S(1,1)SumOfSqs (109.05 ± 4.02 vs. 102.07 ± 1.89, p = 0.02), S(1,1)InvDfMom (0.09 ± 0.02 vs. 0.11 ± 0.01, p = 0.04), S(4,0)Contrast (220.50 ± 27.47 vs. 182.06 ± 4.81, p = 0.04), S(5,0)SumOfSqs (110.26 ± 6.11 vs. 98.90 ± 1.15, p = 0.01), S(5,0)InvDfMom (0.07 ± 0.02 vs. 0.12 ± 0.03, p = 0.04).

Regarding long-term survival, the following texture parameters showed statistical significances: S(1,1)Contrast (survival: 160.84 ± 32.25 vs. non-survival: 191.56 ± 22.83, p = 0.02), S(1,1)SumOfSqs (109.16 ± 3.24 vs. 107.80 ± 5.17, p = 0.01), S(1,1)SumVarnc (275.80 ± 32.57 vs. 239.65 ± 26.03, p = 0.01) (as shown in Fig. 2), S(1,1)DifVarnc (58.18 ± 11.89 vs. 71.45 ± 12.28, p = 0.04), S(2,2)Correlat (0.12 ± 0.15 vs. 0.002 ± 0.07, p = 0.03), S(2,2)SumOfSqs (111.01 ± 4.38 vs. 105.13 ± 5.53, p = 0.01), S(2,2)SumVarnc (247.75 ± 32.07 vs. 210.38 ± 13.16, p = 0.002), S(0,4)InvDfMom (0.08 ± 0.02 vs. 0.07 ± 0.02, p = 0.04), S(0,4)DifVarnc (85.51 ± 17.05 vs. 70.31 ± 13.70, p = 0.04), S(4,4)SumOfSqs (95.30 ± 17.48 vs. 109.68 ± 8.65, p = 0.01), S(5,5)SumOfSqs (86.66 ± 32.61 vs. 110.18 ± 13.48, p = 0.03), S(5,5)SumAverg (62.33 ± 3.09 vs. 66.00 ± 3.62, p = 0.02), S(5,5)DifVarnc (62.96 ± 27.61 vs. 96.72 ± 24.23, p = 0.01), Teta2 (-0.27 ± 0.08 vs. -0.17 ± 0.12, p = 0.01).

3.2 Correlation with tumormarkers

There were moderate statistical significant correlations between several texture features with the tumormarker CEA with Spearman’s correlation coefficients of 0.5 and above:

Skewness (r=-0.66, p = 0.01), S(0,1)DifEntrp (r=-0.611, p = 0.03), S(0,2)DifEntrp (r=-0.655, p = 0.02), S(2,-2)DifVarnc (r=-0.564, p = 0.04), S(3,3)Correlat (r = 0.567, p = 0.04), S(3,3)SumVarnc (r = 0.564, p = 0.04), S(4,-4)InvDfMom (r=-0.578, p = 0.04), S(5,5)Contrast (r = 0.572, p = 0.04), S(5,5)Correlat (r=-0.605, p = 0.03), S(5,5)DifVarnc (r = 0.630, p = 0.02), WavEnLH_s-1 (r=-0.677, p = 0.01), WavEnHL_s-2 (r = 0.732, p = 0.004).

The tumormarker CA 19 − 9 showed statistical significant moderate to good correlations for the following texture parameters with Spearman’s correlation coefficients of 0.5 and above:

_MinNorm (r = 0.543, p = 0.04), S(0,1)InvDfMom (r=-0.66, p = 0.01), S(1,-1)SumVarnc (r=-0.648, p = 0.01), S(1,-1)DifVarnc (r = 0.587, p = 0.03), S(1,-1)DifEntrp (r = 0.591, p = 0.03), S(3,0)SumAverg (r = 0.547, p = 0.04), Vertl_LngREmph (r=-0.560, p = 0.04), Sigma (r = 0.697, p = 0.01, as shown in Fig. 3).

There were strong statistically significant correlations between texture features and the tumormarker AFP with Spearman’s correlation coefficients of 0.75 and above:

S(0,1)AngScMom (r = 0.874, p = 0.005), S(0,1)Entropy (r=-0.838, p = 0.01), S(2,0)Entropy (r=-0.755, p = 0.03), S(0,2)AngScMom (r = 0.838, p = 0.01), S(2,-2)AngScMom (r = 0.778, p = 0.02), S(0,4)SumVarnc (r = 0.790, p = 0.02), S(4,4)AngScMom (r = 0.755, p = 0.03), S(4,4)DifVarnc (r = 0.766, p = 0.03), S(4,-4)AngScMom (r = 0.778, p = 0.02), S(5,-5)SumOfSqs (r=-0.790, p = 0.02).

3.3 Correlation with tumor grading

Correlation of texture parameters with grading showed statistical significant correlation with correlation coefficients of 0.5 and above for:

S(0,1)DifEntrp (r = 0.522, p = 0.01), S(0,2)InvDfMom (r=-0.573, p = 0.01), S(3,0)DifVarnc (r = 0.559, p = 0.01), S(4,0)Contrast (r = 0.508, p = 0.02), S(4,0)DifVarnc (r = 0.568, p = 0.01), S(5,0)SumVarnc (r=-0.587, p = 0.004), S(5,-5)SumAverg (r = 0.519, p = 0.01).

3.4 Correlation with TNM stage

The following texture parameters showed statistical significant corrrelation with T stage with correlation coefficients of 0.5 or above: S(0,5)Contrast (r = 0.527, p = 0.012), S(5,5)DifVarnc (r=-0.499, p = 0.018).

Regarding the N stage texture parameters with N status identified two statistical significant correlations, S(4,-4)Contrast (r = 0.538, p = 0.010), S(4,-4)SumOfSqs (r = 0.579, p = 0.005). There were no statistically significant associations with positive M status.

The present study identified statistically significant associations between texture features derived from CT images of perihilar cholangiocarcinomas with the patient survival. Another key finding is that some texture features showed correlations with grading and T/N-stadium. These findings highlight the possible importance of CT texture analysis to better characterize these kinds of tumors in a non-invasive way. The observed association with the further clinical outcome after surgical treatment might heavily affect clinical management in the future. Interestingly, there were also significant correlations of CT texture parameters with the widely used tumor markers CA-19-9, CEA and AFP.

The morphological CT imaging features of extrahepatic perihilar cholangiocarcinoma are usually presenting with low attenuation pre-contrast and tend to show contrast-enhancement in the portalvenous phase. Sometimes, especially intraductal lesions are not presenting as any visible mass and can be diagnostic challenging. Commonly, proximal dilatation of the bile ducts is present and helps delineating the lesions (20). The clinical outcome of the patients differs significantly according to resectibility (21). Furthermore, poor tumor differentiation has been shown to be a viable predictor of patient long-time survival, a factor that is only poorly reflected by convential CT-imaging. Also, lymph node involvement is an important factor as a further independent predictor of the outcome (22). At the current time, the widely used TNM-staging system and also the Mayo-staging system are enabling preoperative patient prognosis assessment, although the analysis of their prognostic capabilities just have been reported of wielding areas under the curve of only 0.50 and 0.59, respectively (23). Therefore, there is definte need for better characterization of these patients and to provide novel prognostic biomarkets.

The present study only investatigated surgical resected patients. If the radiologists could predict the underlying tumor aggressiveness of the tumor by presurgical imaging, this could improve diagnostic power in the clinical routine as well as potentially reduce invasive procedures. The present results suggest that a promising parameter could be the advanced entropy-based texture feature “SumVarnc”, which might aid in the prediction of patient outcome. This might be accounted for by the well-known possibility of that entropy parameter to quantify the heterogeneity of the tumors, which has been shown to be associated with underlying histopathology features (3).

Our findings are in line with a recent study on the same topic which has been showing the usefulness of radiomics in building multivariate prediction models for early recurrence in a small patient collective (24). Another work group performed a radiomics study on presurgical MRI and found that a combined clinical and radiomics based nomogram performed significantly better in the prediction of 1-, 3- and 5-year survival than the classical TNM-staging system [23]. When comparing the different nomograms, the combined clinical and radiomic approach performed only slightly better than the radiomic nomogram, although with no statistical significant difference. These results are very promising and could prove very helpful in future clinical care employing CT and MRI images (25,26).

Furthermore, a recent PET/CT-study showed the ability of texture analysis to predict pathology data and patient outcome in a cohort with intrahepatic cholangiocarcinoma. Interestingly, in this study it was also shown that also the radiomics of the peritumoral areas had clinical relevance (27). This should be emphasized that not only the texture features of the tumor itself but also the peritumoral region might be of prognostic relevance.

It has to been taken in account though that the stability of texture features between different scanners and reconstruction methods was investigated in many previous studies and is still a challenge for possible future clinical translation (28) Furthermore, the different lesion size has to be considered as a possible factor influencing texture features, as already has been shown in phantom studies (29). Also, the way of contrast media application has to been taken account for, as a recent study has been showing possible differences in some texture parameters between native and contrast enhanced CT scans (30). Yet, in the present analysis only CT images in portalvenous phase were included to reduce heterogeneity. However, texture features derived from CT images are in general considered more reliable compared with texture features from MRI (31). Another recent study investigated the reproducibility of texture features between different segmentations by different operators and also different contrast phases, which could prove a reproducibility ranging from 67–100 % (32) Naturally, aso besides the mentioned challenges of texture analysis in clinical translation, our study is not without some further limitations.

Beyond the inherent limitations of CT texuture analysis, the present study has some limitations to address. As a retrospective study there is a possible known inherent bias. Yet, this was accounted for by blinded reading of the CT to clinical features. Furthermore, as a single center study the patient sample is rather small. Factors such as slight differences in contrast media application, lesion size and localization could have been influencing the texture features analysis.

In conclusion, several texture features derived from CT were associated with survival, tumor stage and tumor markers Therefore, CT texture features could be used as promising novel biomarkers in patients with perihilar cholangiocarcinoma.

Ethical Approval

The study was approved by the local ethical commission board from the University of Leipzig (AZ- EK: 243-14-14072014).

Competing interests

The authors have no competing interests to declare.

Author contribution:

Conception and design: Hans-Jonas Meyer

Administrative support: Daniel Seehofer, Timm Denecke

Provision of study materials or patients: Arsen Sabanov, Jakob Leonhardi, Anne-Kathrin Höhn, Hans-Jonas Meyer

Collection and assembly of data: Jakob Leonhardi, Hans-Jonas Meyer

Data analysis and interpretation: Jakob Leonhardi, Hans-Jonas Meyer

Manuscript writing: All authors

Final approval of manuscript: All authors

Funding

For this study, none of the authors received any funding.

Availability of data and materials

Upon reasonable request.

Incoronato M, Aiello M, Infante T, Cavaliere C, Grimaldi A, Mirabelli P, et al. Radiogenomic Analysis of Oncological Data: A Technical Survey. Int J Mol Sci. 2017 Apr 12;18(4):805.
Wu G, Jochems A, Refaee T, Ibrahim A, Yan C, Sanduleanu S, et al. Structural and functional radiomics for lung cancer. Eur J Nucl Med Mol Imaging. 2021 Nov 11;48(12):3961–74.
Just N. Improving tumour heterogeneity MRI assessment with histograms. Br J Cancer. 2014 Dec 9;111(12):2205–13.
Meyer HJ, Leonhardi J, Höhn AK, Pappisch J, Wirtz H, Denecke T, et al. CT Texture Analysis of Pulmonary Neuroendocrine Tumors—Associations with Tumor Grading and Proliferation. J Clin Med. 2021 Nov 26;10(23):5571.
Meyer HJ, Hamerla G, Höhn AK, Surov A. CT Texture Analysis—Correlations With Histopathology Parameters in Head and Neck Squamous Cell Carcinomas. Front Oncol. 2019 May 28;9.
Meyer HJ, Schob S, Höhn AK, Surov A. MRI Texture Analysis Reflects Histopathology Parameters in Thyroid Cancer – A First Preliminary Study. Transl Oncol. 2017 Dec;10(6):911–6.
Morris LGT, Riaz N, Desrichard A, Şenbabaoğlu Y, Hakimi AA, Makarov V, et al. Pan-cancer analysis of intratumor heterogeneity as a prognostic determinant of survival. Oncotarget. 2016 Mar 1;7(9):10051–63.
Blechacz B, Komuta M, Roskams T, Gores GJ. Clinical diagnosis and staging of cholangiocarcinoma. Nat Rev Gastroenterol Hepatol. 2011 Sep 2;8(9):512–22.
Song C, Kim K, Chie EK, Kim JH, Jang JY, Kim SW, et al. Nomogram Prediction of Survival and Recurrence in Patients With Extrahepatic Bile Duct Cancer Undergoing Curative Resection Followed by Adjuvant Chemoradiation Therapy. International Journal of Radiation Oncology*Biology*Physics. 2013 Nov;87(3):499–504.
Groot Koerkamp B, Wiggers JK, Gonen M, Doussot A, Allen PJ, Besselink MGH, et al. Survival after resection of perihilar cholangiocarcinoma—development and external validation of a prognostic nomogram. Annals of Oncology. 2015 Sep;26(9):1930–5.
Chen HW, Pan AZ, Zhen ZJ, Su SY, Wang JH, Yu SCH, et al. Preoperative Evaluation of Resectability of Klatskin Tumor with 16-MDCT Angiography and Cholangiography. American Journal of Roentgenology. 2006 Jun;186(6):1580–6.
Lee HY, Kim SH, Lee JM, Kim SW, Jang JY, Han JK, et al. Preoperative Assessment of Resectability of Hepatic Hilar Cholangiocarcinoma: Combined CT and Cholangiography with Revised Criteria. Radiology. 2006 Apr;239(1):113–21.
Joo I, Lee JM, Yoon JH. Imaging Diagnosis of Intrahepatic and Perihilar Cholangiocarcinoma: Recent Advances and Challenges. Radiology. 2018 Jul;288(1):7–13.
Strasberg SM. Nomenclature of hepatic anatomy and resections: a review of the Brisbane 2000 system. J Hepatobiliary Pancreat Surg. 2005 Oct 28;12(5):351–5.
Hau HM, Devantier M, Jahn N, Sucher E, Rademacher S, Seehofer D, et al. Impact of Body Mass Index on Tumor Recurrence in Patients Undergoing Liver Resection for Perihilar Cholangiocarcinoma (pCCA). Cancers (Basel). 2021 Sep 24;13(19):4772.
Shimada M, Yamashita Y, Aishima S, Shirabe K, Takenaka K, Sugimachi K. Value of lymph node dissection during resection of intrahepatic cholangiocarcinoma. British Journal of Surgery. 2002 Nov 29;88(11):1463–6.
Strzelecki M, Szczypinski P, Materka A, Klepaczko A. A software tool for automatic classification and segmentation of 2D/3D medical images. Nucl Instrum Methods Phys Res A. 2013 Feb;702:137–40.
Szczypiński PM, Strzelecki M, Materka A, Klepaczko A. MaZda—A software package for image texture analysis. Comput Methods Programs Biomed. 2009 Apr;94(1):66–76.
Fruehwald-Pallamar J, Czerny C, Holzer-Fruehwald L, Nemec SF, Mueller-Mang C, Weber M, et al. Texture-based and diffusion-weighted discrimination of parotid gland lesions on MR images at 3.0 Tesla. NMR Biomed. 2013 Nov;26(11):1372–9.
Chung YE, Kim MJ, Park YN, Choi JY, Pyo JY, Kim YC, et al. Varying Appearances of Cholangiocarcinoma: Radiologic-Pathologic Correlation. RadioGraphics. 2009 May;29(3):683–700.
Spolverato G, Yakoob MY, Kim Y, Alexandrescu S, Marques HP, Lamelas J, et al. The Impact of Surgical Margin Status on Long-Term Outcome After Resection for Intrahepatic Cholangiocarcinoma. Ann Surg Oncol. 2015 Nov 12;22(12):4020–8.
Hau HM, Meyer F, Jahn N, Rademacher S, Sucher R, Seehofer D. Prognostic Relevance of the Eighth Edition of TNM Classification for Resected Perihilar Cholangiocarcinoma. J Clin Med. 2020 Sep 29;9(10):3152.
Sun Z, Sun X, Guo J, Li X, Wang Q, Su N, et al. Prognostic influence for hilar cholangiocarcinoma and comparisons of prognostic values of Mayo staging and TNM staging systems. Medicine. 2022 Dec 9;101(49):e32250.
Qin H, Hu X, Zhang J, Dai H, He Y, Zhao Z, et al. Machine-learning radiomics to predict early recurrence in perihilar cholangiocarcinoma after curative resection. Liver International. 2021 Apr 25;41(4):837–50.
Zhao J, Zhang W, Fan CL, Zhang J, Yuan F, Liu SY, et al. Development and validation of preoperative magnetic resonance imaging-based survival predictive nomograms for patients with perihilar cholangiocarcinoma after radical resection: A pilot study. Eur J Radiol. 2021 May;138:109631.
Zhao J, Zhang W, Zhu Y, Zheng H, Xu L, Zhang J, et al. Development and Validation of Noninvasive < scp > MRI</scp> -Based Signature for Preoperative Prediction of Early Recurrence in Perihilar Cholangiocarcinoma. Journal of Magnetic Resonance Imaging. 2022 Mar 23;55(3):787–802.
Fiz F, Masci C, Costa G, Sollini M, Chiti A, Ieva F, et al. PET/CT-based radiomics of mass-forming intrahepatic cholangiocarcinoma improves prediction of pathology data and survival. Eur J Nucl Med Mol Imaging. 2022 Aug 26;49(10):3387–400.
Mackin D, Fave X, Zhang L, Fried D, Yang J, Taylor B, et al. Measuring Computed Tomography Scanner Variability of Radiomics Features. Invest Radiol. 2015 Nov;50(11):757–65.
Jensen LJ, Kim D, Elgeti T, Steffen IG, Hamm B, Nagel SN. Stability of Radiomic Features across Different Region of Interest Sizes—A CT and MR Phantom Study. Tomography. 2021 Jun 8;7(2):238–52.
Kakino R, Nakamura M, Mitsuyoshi T, Shintani T, Hirashima H, Matsuo Y, et al. Comparison of radiomic features in diagnostic CT images with and without contrast enhancement in the delayed phase for NSCLC patients. Physica Medica. 2020 Jan;69:176–82.
Rekhtman N. Lung neuroendocrine neoplasms: recent progress and persistent challenges. Modern Pathology. 2022 Jan;35:36–50.
Gruzdev IS, Zamyatina KA, Tikhonova VS, Kondratyev EV, Glotov AV, Karmazanovsky GG, et al. Reproducibility of CT texture features of pancreatic neuroendocrine neoplasms. Eur J Radiol. 2020 Dec;133:109371.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

CT texture analysis of perihilar cholangiocarcinoma – associations with tumor grading, tumor markers and clinical outcome

Status:

Version 1

Abstract

Background:

Methods:

Results:

Conclusion:

Figures

1. Introduction

2. Methods

2.1. Study design

2.2. Surgical treatment

2.3. Histopathology and laboratory analyses

2.4. Image analysis

2.5. Texture analysis

2.6. Statistical analysis

3. Results

3.1 Discrimination analysis of survival

3.2 Correlation with tumormarkers

3.3 Correlation with tumor grading

3.4 Correlation with TNM stage

4. Discussion

Declarations

References

Additional Declarations

Status:

Version 1