MRI radiomics for the prediction of recurrence in patients with clinically non-functioning pituitary macroadenomas

被引:33
|
作者
Machado, Leonardo F. [1 ]
Elias, Paula C. L. [3 ]
Moreira, Ayrton C. [3 ]
dos Santos, Antonio C. [4 ]
Murta Junior, Luiz O. [1 ,2 ]
机构
[1] Univ Sao Paulo, Dept Phys, Fac Philosophy Sci & Letters Ribeirao Preto, Ribeirao Preto, SP, Brazil
[2] Univ Sao Paulo, Dept Comp & Math, Fac Philosophy Sci & Letters Ribeirao Preto, Ribeirao Preto, SP, Brazil
[3] Univ Sao Paulo, Clin Hosp, Fac Med Ribeirao Preto, Endocrinol Div, Ribeirao Preto, SP, Brazil
[4] Univ Sao Paulo, Clin Hosp, Fac Med Ribeirao Preto, Endocrinol Div,Ctr Image Sci & Med Phys, Ribeirao Preto, SP, Brazil
关键词
Radiomics; MRI; Pituitary adenoma; Tumor recurrence; Machine learning; TEXTURE ANALYSIS; ADENOMAS; CLASSIFICATION; RADIOTHERAPY; MANAGEMENT; FEATURES; TUMORS; DIAGNOSIS; PATTERN; KI-67;
D O I
10.1016/j.compbiomed.2020.103966
中图分类号
Q [生物科学];
学科分类号
07 ; 0710 ; 09 ;
摘要
Twelve to 66% of patients with clinically non-functioning pituitary adenoma (NFPA) experience tumor recurrence 1-5 years after the first surgery. Nevertheless, there is still no recurrence prediction factor concisely established and reproduced in the literature for NFPA management. The present study evaluates the prognostic value of MRI Radiomics features combined with machine learning models to assess recurrence after the first surgery in patients with clinically non-functioning pituitary adenomas (NFPA). We carried out a retrospective study on 27 patients with NFPA, 10 patients having experienced tumor recurrence after the first surgery and 17 who did not. Preoperative 3D T-1 contrast-enhanced MR images of patients were used to extract up to 255 Radiomics features from two and three-dimensional segmented regions. Additionally, gender, age at first surgery, and the presence of remnant tumor tissue were investigated to find the correlation with NFPA recurrence. Conventional statistics tests were used to evaluate whether the outcome patient groups (stable and recurrent) were different considering each feature individually. Additionally, five well-known machine-learning algorithms were used in combination with Radiomic features to classify recurrent and stable lesions. We found statistical evidence (p < 0.02) for 6 two-dimensional and 13 three-dimensional radiomic features. We achieved accuracies of up to 96.3% for 3D-feature based models and up to 92.6% accuracies for 2D-feature based models. 3D-feature based models achieved better performances using considerably fewer features when compared to 2D-feature based models. We concluded that Radiomics have the potential of NFPA recurrence prediction after the first surgery. Three-dimensional Radiomics have superior discrimination power to predict NFPA recurrence than two-dimensional radiomic features. Finally, the combination of Radiomics with machine-learning algorithms can offer computational models capable of non-invasive, unbiased, and quick assessment that might improve the prediction of NFPA recurrence.
引用
收藏
页数:8
相关论文
共 50 条
  • [41] Vasculogenic Mimicry in Clinically Non-functioning Pituitary Adenomas: a Histologic Study
    Di Michele, Joseph
    Rotondo, Fabio
    Kovacs, Kalman
    Syro, Luis V.
    Yousef, George M.
    Cusimano, Michael D.
    Di Ieva, Antonio
    PATHOLOGY & ONCOLOGY RESEARCH, 2017, 23 (04) : 803 - 809
  • [42] Contemporary Management of Clinically Non-functioning Pituitary Adenomas: A Clinical Review
    AlMalki, Mussa H.
    Ahmad, Maswood M.
    Brema, Imad
    AlDahmani, Khaled M.
    Pervez, Nadeem
    Al-Dandan, Sadeq
    AlObaid, Abdullah
    Beshyah, Salem A.
    CLINICAL MEDICINE INSIGHTS-ENDOCRINOLOGY AND DIABETES, 2020, 13
  • [43] A clinicopathological and immunohistochemical study of clinically non-functioning pituitary adenomas: A single institutional experience
    Rishi, Arvind
    Sharma, Mehar C.
    Sarkar, Chitra
    Jain, Deepali
    Singh, Manmohan
    Mahapatra, Ashok Kumar
    Mehta, Veer Singh
    Das, Tapos Kumar
    NEUROLOGY INDIA, 2010, 58 (03) : 418 - 423
  • [44] Prevalence, Clinical Features, and Natural History of Incidental Clinically Non-Functioning Pituitary Adenomas
    Iglesias, Pedro
    Arcano, Karina
    Trivino, Vanessa
    Garcia-Sancho, Paula
    Jose Diez, Juan
    Villabona, Carles
    Cordido, Fernando
    HORMONE AND METABOLIC RESEARCH, 2017, 49 (09) : 654 - 659
  • [45] Non-functioning pituitary adenomas
    Greenman, Yona
    Stern, Naftali
    BEST PRACTICE & RESEARCH CLINICAL ENDOCRINOLOGY & METABOLISM, 2009, 23 (05) : 625 - 638
  • [46] Treatment with octreotide LAR in clinically non-functioning pituitary adenoma: results from a case-control study
    Fusco, Alessandra
    Giampietro, Antonella
    Bianchi, Antonio
    Cimino, Vincenzo
    Lugli, Francesca
    Piacentini, Serena
    Lorusso, Margherita
    Tofani, Anna
    Perotti, Germano
    Lauriola, Libero
    Anile, Carmelo
    Maira, Giulio
    Pontecorvi, Alfredo
    De Marinis, Laura
    PITUITARY, 2012, 15 (04) : 571 - 578
  • [47] Perioperative cortisol can predict hypothalamus-pituitary-adrenal status in clinically non-functioning pituitary adenomas
    Cozzi, R.
    Lasio, G.
    Cardia, A.
    Felisati, G.
    Montini, M.
    Attanasio, R.
    JOURNAL OF ENDOCRINOLOGICAL INVESTIGATION, 2009, 32 (05) : 460 - 464
  • [48] Non-invasive radiomics approach potentially predicts non-functioning pituitary adenomas subtypes before surgery
    Zhang, Shuaitong
    Song, Guidong
    Zang, Yali
    Jia, Jian
    Wang, Chao
    Li, Chuzhong
    Tian, Jie
    Dong, Di
    Zhang, Yazhuo
    EUROPEAN RADIOLOGY, 2018, 28 (09) : 3692 - 3701
  • [49] Non-functioning pituitary tumors: 2012 update
    Camara Gomez, Rosa
    ENDOCRINOLOGIA Y NUTRICION, 2014, 61 (03): : 160 - 170
  • [50] Management of non-functioning pituitary adenomas: surgery
    Penn, David L.
    Burke, William T.
    Laws, Edward R.
    PITUITARY, 2018, 21 (02) : 145 - 153