Synopsis

Keywords: Body: Pelvis

Motivation:T2WI, i.e. traditional morphological image diagnosis is good at ‘present’ histopathological diagnosis. On the other hand, recent texture, i.e. radiomics analysis can provide invisible information and make ‘future’ diagnosis including prognosis and treatment response.
Goal(s): Clinical radiologists need to know the pros and cons of both morphological diagnosis and texture analysis and know how to get the necessary information when we need it.
Approach: -
Results: -
Impact: -

The fundamental role of MR imaging about ovarian tumor is to distinguish malignant tumor from benign tumors. The majority of ovarian tumors are benign and can be effectively addressed through conservative management. The preoperative accurate diagnosis helps in avoiding unnecessary surgical procedures and reducing healthcare expenses [1]. Ultimately, histopathological diagnosis is required to construct proper treatment strategie O-RADS MRI s. To achieve this objective, morphological information, including shape, signal intensity, degree of enhancement etc., has been used by combining several MRI sequences. Recent risk stratification system of ovarian lesion, called the ovarian-adnexal reporting and data system on MRI (), also based on the morphological diagnosis. Combining several morphological features, ORADS score indicate their likelihood of malignancy [2]. Their diagnostic results showed excellent ability, the area under the receiver operating characteristic curve (AUC) was 0.961 among experienced readers [3]. Even though without contrast enhancement only using T1, T2 and DWI, differentiation between malignant and benign tumor was capable in high precision in AUC of 0.93 [4]. As is well known, there are characteristic imaging findings depending on each ovarian tumors, such as: solid tumor with low signal intensity both in T2WI and DWI may indicate fibroma, multiple cystic tumors may be mucinous cystadenoma, and hemorrhagic cystic tumor with mural nodule might be clear cell carcinoma etc. The knowledge of those characteristic image findings may lead to histopathological diagnosis. The shortage of morphological diagnosis is qualitative analysis containing individual bias depending on the experience. Also, it is difficult to get other information other than histopathological diagnosis, such as predicting treatment response or prognosis. Radiomics has great potential that provide variety of information that are beyond visual assessment from the same diagnostic images. It is expected to establish connections between quantitative data extracted from imaging studies and clinical information, with the goal of facilitating evidence-based clinical decision-making [5]. Texture analysis is a part of radiomics and allows quantitative evaluation of the degree of image heterogeneity within the tumor. Using these analyses, many researchers tried to predict prognosis and treatment response by the quantitative information extracted from CT or MRI. The imitation of texture analysis is that analytic methods are very difficult to understand for clinical radiologists and to use clinically. As for prediction of prognosis, Lu et al. developed radiomic prognostic vector of the primary ovarian tumor and made their original model that indicate good correlation with prognosis of high grade serous carcinoma (HGSC) patients [6]. Regarding the prediction of treatment response, Boehm KM et al. made a combined machine-learning model of CT, histopathological specimens, genomics and clinical data to improve prediction of treatment response of HGSC. They found tumor nuclear size on staining with hematoxylin and eosin and omental texture on contrast-enhanced (CE)-CT, associated with prognosis. Highly dense omental implants are an adverse prognostic factor and omental texture measured by autocorrelation may reflect intratumoral heterogeneity [7]. Another report by Crispin-Ortuzar M et al. tried to predict the response of HGSOC patients to neoadjuvant chemotherapy using CE-CT. They integrated baseline clinical, blood-based, and radiomic biomarkers and made a machine learning model trained to predict the change in total disease volume. Only the models that included Radiomic features could produce response scores that were significantly correlated with the observed volume response [8]. Radiomics-clinical nomograms for prognosis prediction was constructed by Wang T et al. based on MR images. The T2WI radiomic-clinical nomogram achieved a favorable prediction performance with an AUC of 0.818 in validation cohort [9]. As shown above, radiomics can provide invisible information and could help to predict therapeutic outcome before treatment or provide methods for developing new biomarker-based clinical trials [8, 9]. Histopathological classification by radiomics is also tried in several articles, but not enough to visual morphological assessment. For example, Qian L. et al. tried to distinguish Type I tumor from Type II tumor, and Ye R et al. showed differentiation of borderline tumor from malignant tumor by 3D texture [10, 11]. Therefore, T2, i.e. morphological diagnosis can be good at ‘present’ histopathological diagnosis, and texture, i.e. radiomic analysis can provide invisible and ‘future’ diagnosis including prognosis and treatment response. Accordingly, clinical radiologists need to know the pros and cons of both methods and know how to extract necessary information adequately when needed.

Acknowledgements

No acknowledgement found.

References

[1] S. Nougaret, C. McCague, H. Tibermacine, H.A. Vargas, S. Rizzo, E. Sala, Radiomics and radiogenomics in ovarian cancer: a literature review, Abdom Radiol (NY), 46 (2021) 2308-2322. [2] I. Thomassin-Naggara, Y. Dabi, M. Florin, A. Saltel-Fulero, L. Manganaro, M. Bazot, L. Razakamanantsoa, O-RADS MRI SCORE: An Essential First-Step Tool for the Characterization of Adnexal Masses, J Magn Reson Imaging, DOI 10.1002/jmri.28947(2023). [3] P.N. Pereira, L.O. Sarian, A. Yoshida, K.G. Araujo, A.C.B. Silva, R.H. de Oliveira Barros, R.M. Jales, S. Derchain, Improving the performance of IOTA simple rules: sonographic assessment of adnexal masses with resource-effective use of a magnetic resonance scoring (ADNEX MR scoring system), Abdom Radiol (NY), 45 (2020) 3218-3229. [4] H. Sahin, C. Panico, S. Ursprung, V. Simeon, P. Chiodini, A. Frary, B. Carmo, J. Smith, S. Freeman, M. Jimenez-Linan, H. Bolton, K. Haldar, J.E. Ang, C. Reinhold, E. Sala, H. Addley, Non-contrast MRI can accurately characterize adnexal masses: a retrospective study, Eur Radiol, 31 (2021) 6962-6973. [5] S. Rizzo, F. Botta, S. Raimondi, D. Origgi, C. Fanciullo, A.G. Morganti, M. Bellomi, Radiomics: the facts and the challenges of image analysis, Eur Radiol Exp, 2 (2018) 36. [6] H. Lu, M. Arshad, A. Thornton, G. Avesani, P. Cunnea, E. Curry, F. Kanavati, J. Liang, K. Nixon, S.T. Williams, M.A. Hassan, D.D.L. Bowtell, H. Gabra, C. Fotopoulou, A. Rockall, E.O. Aboagye, A mathematical-descriptor of tumor-mesoscopic-structure from computed-tomography images annotates prognostic- and molecular-phenotypes of epithelial ovarian cancer, Nat Commun, 10 (2019) 764. [7] K.M. Boehm, E.A. Aherne, L. Ellenson, I. Nikolovski, M. Alghamdi, I. Vazquez-Garcia, D. Zamarin, K. Long Roche, Y. Liu, D. Patel, A. Aukerman, A. Pasha, D. Rose, P. Selenica, P.I. Causa Andrieu, C. Fong, M. Capanu, J.S. Reis-Filho, R. Vanguri, H. Veeraraghavan, N. Gangai, R. Sosa, S. Leung, A. McPherson, J. Gao, M.M. Consortium, Y. Lakhman, S.P. Shah, Multimodal data integration using machine learning improves risk stratification of high-grade serous ovarian cancer, Nat Cancer, 3 (2022) 723-733. [8] M. Crispin-Ortuzar, R. Woitek, M.A.V. Reinius, E. Moore, L. Beer, V. Bura, L. Rundo, C. McCague, S. Ursprung, L. Escudero Sanchez, P. Martin-Gonzalez, F. Mouliere, D. Chandrananda, J. Morris, T. Goranova, A.M. Piskorz, N. Singh, A. Sahdev, R. Pintican, M. Zerunian, N. Rosenfeld, H. Addley, M. Jimenez-Linan, F. Markowetz, E. Sala, J.D. Brenton, Integrated radiogenomics models predict response to neoadjuvant chemotherapy in high grade serous ovarian cancer, Nat Commun, 14 (2023) 6756. [9] T. Wang, H. Wang, Y. Wang, X. Liu, L. Ling, G. Zhang, G. Yang, H. Zhang, MR-based radiomics-clinical nomogram in epithelial ovarian tumor prognosis prediction: tumor body texture analysis across various acquisition protocols, J Ovarian Res, 15 (2022) 6. [10] L. Qian, J. Ren, A. Liu, Y. Gao, F. Hao, L. Zhao, H. Wu, G. Niu, MR imaging of epithelial ovarian cancer: a combined model to predict histologic subtypes, Eur Radiol, 30 (2020) 5815-5825. [11] R. Ye, S. Weng, Y. Li, C. Yan, J. Chen, Y. Zhu, L. Wen, Texture Analysis of Three-Dimensional MRI Images May Differentiate Borderline and Malignant Epithelial Ovarian Tumors, Korean J Radiol, 22 (2021) 106-117.