You're using an outdated browser. Please upgrade to a modern browser for the best experience.
Machine Learning for Differentiating Bone Malignancy on Imaging: Comparison
View Latest Version
Please note this is a comparison between Version 2 by Lindsay Dong and Version 1 by Wilson Ong.

Distinguishing between benign vs. malignant bone lesions is often difficult on imaging. Many bone lesions are infrequent or rarely seen, and often only specialist radiologists have sufficient expertise to provide an accurate diagnosis. In addition, some benign bone tumours may exhibit potentially aggressive features that mimic malignant bone tumours, making the diagnosis even more difficult. The rapid development of artificial intelligence (AI) techniques has led to remarkable progress in image-recognition tasks, including the classification and characterization of various tumours. The use of AI to discriminate bone lesions on imaging has achieved a relatively good performance in various imaging modalities, with high sensitivity, specificity, and accuracy for distinguishing between benign vs. malignant lesions in several cohort studies. However, further research is necessary to test the clinical performance of these algorithms before they can be facilitated and integrated into routine clinical practice.

  • deep learning
  • machine learning
  • artificial intelligence
  • bone malignancy

1. Introduction

Differentiating between benign vs. malignant bone tumours is critical for clinical decision-making and treatment planning [1,2][1][2]. Routine imaging investigations for evaluating bone lesions include radiographs (X-ray), computed tomography (CT), positron-emission tomography combined with computed tomography (PET/CT), and magnetic resonance imaging (MRI) [3,4][3][4]. Conventional radiography remains a key initial imaging modality and is still an optimal technique for evaluating primary bone tumours [5,6][5][6]. It is relatively inexpensive and allows for a clear visual assessment of lesion location, margins, internal lesion matrix, and any associated periosteal reaction [7]. Along with the patient’s age, these radiographic details are often sufficient to provide a reasonable list of differential diagnoses [8]. However, radiographs are often limited by superimpositions, incomplete visualizations of bone cortex destruction, and inadequate assessments of adjacent soft tissue involvement [9,10,11][9][10][11]. Furthermore, diagnosis of bone tumours using imaging can be complicated by other factors such as the presence of pathological fractures, which may result in benign bone lesions having potentially aggressive features that mimic malignant bone tumours [12,13][12][13].
Emerging Artificial Intelligence (AI) tools continue to demonstrate remarkable progress in medical imaging applications, especially in the field of oncology [36][14]. These applications include cancer screening and diagnosis [37[15][16][17][18],38,39,40], diagnosis and classification [41[19][20][21][22],42,43,44], predicting prognosis and treatment response [45[23][24][25][26][27],46,47,48,49], automated segmentation [50[28][29][30][31][32],51,52,53,54], and radiology-pathology correlation (radiogenomics) [55,56,57,58][33][34][35][36]. In particular, within the field of diagnosis and classification, the ability of AI models to classify benign vs. malignant tumours has been shown to achieve high accuracy, sensitivity, and specificity in various organs, such as in the case of breast [59[37][38][39],60,61], prostate [62[40][41],63], lung [38,64[16][42][43][44],65,66], and brain lesions [67,68][45][46]. 

2. Machine Learning for Differentiating Bone Malignancy on Imaging

2.1. Machine Learning on Conventional Radiographs

Correctly classifying bone tumours on conventional radiographs is vital for clinical decision-making and guiding subsequent management [120][47]. However, this is often difficult, especially in places where there is a shortage of subspecialty radiology expertise. Moreover, many bone lesions are uncommon entities, and often only a few specialist radiologists have sufficient experience to diagnose them accurately [10,79][10][48]. In clinical practice, most radiologists rely on image pattern recognition to distinguish between benign and malignant lesions on radiographs, which is subject to bias and can sometimes lead to an erroneous interpretation [121,122][49][50]. Some of these common radiological features include location, cortical destruction, periostitis, lesion orientation or alignment, and the zone of transition between the lesion and the surrounding bone [123,124,125,126][51][52][53][54]. However, some benign bone lesions may demonstrate one or more aggressive features which may confound the distinction [127,128][55][56].
Malignant bone lesions are often difficult to differentiate from other aggressive disease processes, including inflammation and infection. Ewing’s sarcoma, an aggressive malignant tumour occurring in children, is a typical example [130][57]. Differentiating this entity from acute osteomyelitis is often difficult, even by trained musculoskeletal radiologists, due to its similar clinical and radiological features [131,132,133][58][59][60]. Consalvo et al. [97][61] developed an artificial intelligence algorithm which was able to leverage radiographic features to distinguish between Ewing sarcoma and acute osteomyelitis, achieving an accuracy of up to 94.4% on their validation set and 90.6% on a held-out test set. Although this study is limited by a small sample size, requiring the use of cross-validation and loss weighting to achieve statistical significance, it demonstrates the potential feasibility of AI techniques for differentiating infective bony lesions from malignant bone lesions on routine radiographs.

2.2. Machine Learning on Computed Tomography (CT) Imaging

Incidentally detected dense (sclerotic) lesions are a common occurrence on CT examinations in clinical practice [134][62]. The ability to distinguish a benign sclerotic lesion, such as an enostosis (bone island), vs. a malignant lesion, such as an osteoblastic metastasis, is crucial as it affects the treatment strategy and patient prognosis [135][63]. For this task, radiomics-based random forest models were created by Hong et al. [85][64] and achieved an AUC of up to 0.96 and an accuracy of up to 86.0% in the test sets, which was comparable to two experienced radiologists (AUCs of 0.95–0.96) and even higher compared to a radiologist in training (AUC 0.88, p = 0.03). Along with the extraction of 1218 radiomics features, the autscholars showed that a model utilizing attenuation and shape-related features achieved the highest AUC, which had been postulated in several prior studies [136,137,138,139,140][65][66][67][68][69]. 

Besides differentiating between benign vs. malignant bony lesions, radiomics models can also be used to differentiate between a variety of tumour matrix types with high performance. A deep convolutional neural network (CNN) created by Y. Li et al. [81][70] was able to further classify benign and malignant bone lesions into cartilaginous or osteogenic tumours using a multi-channel enhancement strategy for image processing to improve accuracy, achieving a top-1 error rate of only 0.25. 

2.3. Machine Learning on Magnetic Resonance Imaging (MRI)

MRI plays a key role in aiding clinicians in discriminating between benign vs. primary malignant or metastatic bone tumours [143][71]. The conventional pulse sequences [20,144][72][73], diffusion-weighted imaging (DWI) [145,146][74][75] with matched apparent diffusion coefficient (ADC) maps [147,148][76][77] as well as dynamic contrast sequences (DCE) [149,150][78][79] can predict potential malignancy with good reliability. However, some imaging features of benign vs. malignant bone lesions can overlap and this makes formulating a differential diagnosis challenging [151][80]. Machine learning techniques using radiomic features on MRI have been shown to have high performance for predicting benign vs. malignant bone lesions. Specific to cartilaginous bone lesions, machine learning has been studied for the differentiation of various grades of chondrosarcoma. Conventional chondrosarcoma is usually divided into three categories based on pathology, where grade 1, also known as atypical cartilaginous tumours (ACTs), usually have an indolent biologic nature, whereas grades 2–3 (high-grade chondrosarcoma) are malignant bone tumours [152][81] with metastatic potential and a high recurrence rate following surgical resection [153][82]. Discrepancies in correct tumour grading are widespread even among experienced radiologists and pathologists, secondary to overlapping imaging and histological features, and it is for this reason that more accurate diagnostic aids are required [154,155][83][84]. The use of intravenous contrast media (gadolinium-based) for the evaluation of bone tumours has been shown to add some specificity in tissue characterization [156[85][86],157], although the main advantages are for the accurate evaluation of tumour extent for biopsy and treatment planning and to assess for recurrence [158,159,160][87][88][89]. Interestingly, a study by Eweje et al. [72][90] demonstrated that a deep learning model was able to achieve a performance similar to that of expert radiologists for classifying bone lesions at various skeletal locations without using post-contrast T1-weighted sequences, which were made available to the radiologists. The model had an accuracy of 76% vs. 73% for radiologists (p = 0.7) for classifying benign (which includes intermediate as per WHO classification criteria) vs. malignant bone lesions. This preliminary study shows the potential utility of machine learning for the accurate diagnosis of bone tumours without requiring the administration of gadolinium-based MRI contrast media. This could be advantageous, especially in patients who have contraindications to gadolinium-based contrast (due to renal impairment or allergy) or in children with pain-related anxiety that is secondary to the placement of an intravenous cannula and the uncertain long-term implications of gadolinium deposition in children [161,162][91][92]. Further larger studies are required to show if machine learning using a combination of non-contrast and contrast-enhanced imaging has an advantage over non-contrast imaging alone.

2.4. Machine Learning on Positiron Emission Tomography with CT (PET/CT) Imaging

PET/CT imaging is a widely used modality to differentiate malignant vs. benign tumours in a host of organ systems [163,164,165][93][94][95]. The most common radiotracer used in PET/CT imaging is 2-deoxy-2-18F-fluoro-β-D-glucose (18F-FDG), an analogue of glucose, with the concentrations of radiotracer accumulation in PET/CT image proportional to the metabolic activity of tissues concerning the underlying glucose accumulation and metabolism [166][96]. Fluorine 18–Sodium Fluoride (18F–NaF) is another radiotracer used more specifically for bone imaging in PET/CT, with the uptake proportional to blood flow in the bone and osseous remodeling [167,168,169][97][98][99]. Increased 18F-NaF bone uptake can be seen in an abnormal bone that undergoes higher remodeling, such as in osteoblastic or osteolytic processes and is used to differentiate various pathologies [170][100]. For the musculoskeletal system, the utility of PET/CT in distinguishing between malignant and benign bone tumours has also been widely studied and proved effective [171,172,173][101][102][103].

To improve the current diagnostic efficacy of PET/CT interpretation, Fan et al. [184][104] utilized texture analysis with SUVmax to construct radiomics models to distinguish between benign vs. malignant bone lesions. Texture analysis extracts information, regarding the relationship between adjacent voxels or pixels, and assesses for inhomogeneity, which can then be used to predict the likelihood of benign vs. malignant bone lesions. [185,186][105][106]. By incorporating partial texture features along with SUVmax, the developed classification model (using logistic regression) achieved an accuracy of 87.5% compared to 84.3% for nuclear medicine physicians (p = 0.03) in differentiating spinal osseous metastases from benign osseous lesions with high SUVmax values.

2.5. Potential Clinical Impact and Applications

There is significant clinical value in the ability of machine learning to differentiate between benign vs. malignant bone lesions. A retrospective study by Stacy et al. [187][107] found that at least a third of patients with bone lesions referred to orthopedic oncology in a year had images that were diagnosed by radiologists as characteristic of benign tumours or non-malignant entities that did not require follow-up or referrals. Accurate AI models for the characterization of osseous lesions could therefore potentially reduce the rate of unnecessary specialist referrals and follow-up, reducing the associated healthcare costs and patient anxiety regarding a possible cancer diagnosis. Moreover, more accurate characterizations of bone lesions would be valuable for radiologists to identify high-risk bone lesions that will benefit from biopsy to rule in malignancy with greater certainty. Unnecessary biopsies of benign bone lesions leave patients at risk of post-procedural complications, and hasty biopsy planning can increase the risk of misdiagnosis, creating unwarranted patient stress [188,189][108][109]. Biopsies can also be non-diagnostic in up to 30% of bone lesions, which requires repeat biopsies and an associated higher risk of complications [190][110]. A robust AI model that can identify benign bone lesions with a high specificity could help reduce the rate of unnecessary biopsies. This would be especially helpful for bone tumour subtypes which demonstrate similar imaging and histology features for their benign and malignant counterparts, for example, cartilaginous tumours including enchondroma (benign) vs. chondrosarcoma [135,165][63][95]. On top of differentiating bone lesions into benign vs. malignant, machine learning methods can also help differentiate lesions secondary to post-treatment change from residual or recurrent malignant disease. This is a crucial diagnostic challenge as the treatment for the two entities are vastly different, and accurate diagnosis is also important to prevent unnecessary invasive biopsy and/or chemoradiotherapy. However, radiological evaluation of bone tumours after treatment can be quite difficult [191,192][111][112].  The use of multiple imaging modalities to evaluate bone tumours is known to improve the accuracy of diagnosis [3]. Machine learning models built using different modalities (multimodal) have also been shown to improve diagnostic performance [193][113]. In breast radiology, Antropova et al. [194][114] came out with a CNN method involving fusion-based classification using dynamic contrast enhanced-MRI, full-field digital mammography, and ultrasound. This method outperformed conventional CNN-based and CADx-based classifiers, with an AUC of 0.90.

References

  1. Ladd, L.M.; Roth, T.D. Computed Tomography and Magnetic Resonance Imaging of Bone Tumors. Semin. Roentgenol. 2017, 52, 209–226.
  2. Piperkova, E.; Mikhaeil, M.; Mousavi, A.; Libes, R.; Viejo-Rullan, F.; Lin, H.; Rosen, G.; Abdel-Dayem, H. Impact of PET and CT in PET/CT studies for staging and evaluating treatment response in bone and soft tissue sarcomas. Clin. Nucl. Med. 2009, 34, 146–150.
  3. Goyal, N.; Kalra, M.; Soni, A.; Baweja, P.; Ghonghe, N.P. Multi-modality imaging approach to bone tumors—State-of-the art. J. Clin. Orthop. Trauma. 2019, 10, 687–701.
  4. Hapani, H.; Kalola, J.; Hapani, J. Comparative role of CT scan and MR imaging in primary malignant bone tumors. IOSR J. Dent. Med. Sci. 2014, 13, 29–35.
  5. Berquist, T.; Dalinka, M.; Alazraki, N.; Daffner, R.; DeSmet, A.; el-Khoury, G.; Goergen, T.; Keats, T.; Manaster, B.; Newberg, A. Bone tumors. American college of radiology. ACR appropriateness criteria. Radiology 2000, 215, 261–264.
  6. Costelloe, C.M.; Rohren, E.M.; Madewell, J.E.; Hamaoka, T.; Theriault, R.L.; Yu, T.-K.; Lewis, V.O.; Ma, J.; Stafford, R.J.; Tari, A.M.; et al. Imaging bone metastases in breast cancer: Techniques and recommendations for diagnosis. Lancet Oncol. 2009, 10, 606–614.
  7. Karanian, M.; Coindre, J.M. Fourth edition of WHO classification tumours of soft tissue. Ann. Pathol. 2015, 35, 71–85.
  8. Vanel, D. General principles of imaging. In Diagnosis of Musculoskeletal Tumors and Tumor-like Conditions: Clinical, Radiological and Histological Correlations—The Rizzoli Case Archive; Picci, P., Manfrini, M., Donati, D.M., Gambarotti, M., Righi, A., Vanel, D., Dei Tos, A.P., Eds.; Springer International Publishing: Cham, Switzerland, 2020; pp. 27–30.
  9. Huang, P.-Y.; Wu, P.-K.; Chen, C.-F.; Lee, F.-T.; Wu, H.-T.; Liu, C.-L.; Chen, T.-H.; Chen, W.-M. Osteomyelitis of the femur mimicking bone tumors: A review of 10 cases. World J. Surg. Oncol. 2013, 11, 283.
  10. Gerber, E.; Said-Hartley, Q.; Gamieldien, R.; Hartley, T.; Candy, S. Accuracy of plain radiographs in diagnosing biopsy-proven malignant bone lesions. SA J. Radiol. 2019, 23, 1768.
  11. Priolo, F.; Cerase, A. The current role of radiography in the assessment of skeletal tumors and tumor-like lesions. Eur. J. Radiol. 1998, 27 (Suppl. S1), S77–S85.
  12. Picci, P. Epidemiology of Bone Lesions. In Diagnosis of Musculoskeletal Tumors and Tumor-like Conditions: Clinical, Radiological and Histological Correlations—The Rizzoli Case Archive; Picci, P., Manfrini, M., Donati, D.M., Gambarotti, M., Righi, A., Vanel, D., Dei Tos, A.P., Eds.; Springer International Publishing: Cham, Switzerland, 2020; pp. 3–9.
  13. Wu, J.S.; Hochman, M.G. Bone Tumors: A Practical Guide to Imaging; Springer Science & Business Media: Berlin/Heidelberg, Germany, 2012.
  14. Tran, K.A.; Kondrashova, O.; Bradley, A.; Williams, E.D.; Pearson, J.V.; Waddell, N. Deep learning in cancer diagnosis, prognosis and treatment selection. Genome Med. 2021, 13, 152.
  15. Liu, B.; Chi, W.; Li, X.; Li, P.; Liang, W.; Liu, H.; Wang, W.; He, J. Evolving the pulmonary nodules diagnosis from classical approaches to deep learning-aided decision support: Three decades’ development course and future prospect. J. Cancer Res. Clin. Oncol. 2020, 146, 153–185.
  16. Li, K.; Liu, K.; Zhong, Y.; Liang, M.; Qin, P.; Li, H.; Zhang, R.; Li, S.; Liu, X. Assessing the predictive accuracy of lung cancer, metastases, and benign lesions using an artificial intelligence-driven computer aided diagnosis system. Quant Imaging Med. Surg. 2021, 11, 3629–3642.
  17. Raya-Povedano, J.L.; Romero-Martín, S.; Elías-Cabot, E.; Gubern-Mérida, A.; Rodríguez-Ruiz, A.; Álvarez-Benito, M. AI-based Strategies to Reduce Workload in Breast Cancer Screening with Mammography and Tomosynthesis: A Retrospective Evaluation. Radiology 2021, 300, 57–65.
  18. Graewingholt, A.; Duffy, S. Retrospective comparison between single reading plus an artificial intelligence algorithm and two-view digital tomosynthesis with double reading in breast screening. J. Med. Screen. 2021, 28, 365–368.
  19. Chakrabarty, S.; Sotiras, A.; Milchenko, M.; LaMontagne, P.; Hileman, M.; Marcus, D. MRI-based Identification and Classification of Major Intracranial Tumor Types by Using a 3D Convolutional Neural Network: A Retrospective Multi-institutional Analysis. Radiol. Artif. Intell. 2021, 3, e200301.
  20. Deepak, S.; Ameer, P.M. Brain tumor classification using deep CNN features via transfer learning. Comput. Biol. Med. 2019, 111, 103345.
  21. Díaz-Pernas, F.J.; Martínez-Zarzuela, M.; Antón-Rodríguez, M.; González-Ortega, D. A Deep Learning Approach for Brain Tumor Classification and Segmentation Using a Multiscale Convolutional Neural Network. Healthcare 2021, 9, 153.
  22. Dmitriev, K.; Kaufman, A.E.; Javed, A.A.; Hruban, R.H.; Fishman, E.K.; Lennon, A.M.; Saltz, J.H. Classification of Pancreatic Cysts in Computed Tomography Images Using a Random Forest and Convolutional Neural Network Ensemble. Med. Image Comput. Comput. Assist. Interv. 2017, 10435, 150–158.
  23. Massafra, R.; Fanizzi, A.; Amoroso, N.; Bove, S.; Comes, M.C.; Pomarico, D.; Didonna, V.; Diotaiuti, S.; Galati, L.; Giotta, F.; et al. Analyzing breast cancer invasive disease event classification through explainable artificial intelligence. Front. Med. 2023, 10, 1116354.
  24. Du, R.; Lee, V.H.; Yuan, H.; Lam, K.O.; Pang, H.H.; Chen, Y.; Lam, E.Y.; Khong, P.L.; Lee, A.W.; Kwong, D.L.; et al. Radiomics Model to Predict Early Progression of Nonmetastatic Nasopharyngeal Carcinoma after Intensity Modulation Radiation Therapy: A Multicenter Study. Radiol. Artif. Intell. 2019, 1, e180075.
  25. Bibault, J.E.; Giraud, P.; Housset, M.; Durdux, C.; Taieb, J.; Berger, A.; Coriat, R.; Chaussade, S.; Dousset, B.; Nordlinger, B.; et al. Deep Learning and Radiomics predict complete response after neo-adjuvant chemoradiation for locally advanced rectal cancer. Sci. Rep. 2018, 8, 12611.
  26. Kao, Y.S.; Hsu, Y. A Meta-Analysis for Using Radiomics to Predict Complete Pathological Response in Esophageal Cancer Patients Receiving Neoadjuvant Chemoradiation. Vivo 2021, 35, 1857–1863.
  27. DiCenzo, D.; Quiaoit, K.; Fatima, K.; Bhardwaj, D.; Sannachi, L.; Gangeh, M.; Sadeghi-Naini, A.; Dasgupta, A.; Kolios, M.C.; Trudeau, M.; et al. Quantitative ultrasound radiomics in predicting response to neoadjuvant chemotherapy in patients with locally advanced breast cancer: Results from multi-institutional study. Cancer Med. 2020, 9, 5798–5806.
  28. Lin, M.; Momin, S.; Lei, Y.; Wang, H.; Curran, W.J.; Liu, T.; Yang, X. Fully automated segmentation of brain tumor from multiparametric MRI using 3D context deep supervised U-Net. Med. Phys. 2021, 48, 4365–4374.
  29. Primakov, S.P.; Ibrahim, A.; van Timmeren, J.E.; Wu, G.; Keek, S.A.; Beuque, M.; Granzier, R.W.Y.; Lavrova, E.; Scrivener, M.; Sanduleanu, S.; et al. Automated detection and segmentation of non-small cell lung cancer computed tomography images. Nat. Commun. 2022, 13, 3423.
  30. Chang, C.Y.; Buckless, C.; Yeh, K.J.; Torriani, M. Automated detection and segmentation of sclerotic spinal lesions on body CTs using a deep convolutional neural network. Skelet. Radiol. 2022, 51, 391–399.
  31. Goehler, A.; Harry Hsu, T.M.; Lacson, R.; Gujrathi, I.; Hashemi, R.; Chlebus, G.; Szolovits, P.; Khorasani, R. Three-Dimensional Neural Network to Automatically Assess Liver Tumor Burden Change on Consecutive Liver MRIs. J. Am. Coll. Radiol. 2020, 17, 1475–1484.
  32. Anderson, B.M.; Rigaud, B.; Lin, Y.-M.; Jones, A.K.; Kang, H.C.; Odisio, B.C.; Brock, K.K. Automated segmentation of colorectal liver metastasis and liver ablation on contrast-enhanced CT images. Front. Oncol. 2022, 12, 886517.
  33. Fathi Kazerooni, A.; Bagley, S.J.; Akbari, H.; Saxena, S.; Bagheri, S.; Guo, J.; Chawla, S.; Nabavizadeh, A.; Mohan, S.; Bakas, S.; et al. Applications of Radiomics and Radiogenomics in High-Grade Gliomas in the Era of Precision Medicine. Cancers 2021, 13, 5921.
  34. Saxena, S.; Jena, B.; Gupta, N.; Das, S.; Sarmah, D.; Bhattacharya, P.; Nath, T.; Paul, S.; Fouda, M.M.; Kalra, M.; et al. Role of Artificial Intelligence in Radiogenomics for Cancers in the Era of Precision Medicine. Cancers 2022, 14, 2860.
  35. Ren, M.; Yang, H.; Lai, Q.; Shi, D.; Liu, G.; Shuang, X.; Su, J.; Xie, L.; Dong, Y.; Jiang, X. MRI-based radiomics analysis for predicting the EGFR mutation based on thoracic spinal metastases in lung adenocarcinoma patients. Med. Phys. 2021, 48, 5142–5151.
  36. Darvish, L.; Bahreyni-Toossi, M.-T.; Roozbeh, N.; Azimian, H. The role of radiogenomics in the diagnosis of breast cancer: A systematic review. Egypt. J. Med. Hum. Genet. 2022, 23, 99.
  37. Zhou, Y.; Feng, B.-J.; Yue, W.-W.; Liu, Y.; Xu, Z.-F.; Xing, W.; Xu, Z.; Yao, J.-C.; Wang, S.-R.; Xu, D. Differentiating non-lactating mastitis and malignant breast tumors by deep-learning based AI automatic classification system: A preliminary study. Front. Oncol. 2022, 12, 997306.
  38. Kayode, A.A.; Akande, N.O.; Adegun, A.A.; Adebiyi, M.O. An automated mammogram classification system using modified support vector machine. Med. Devices 2019, 12, 275–284.
  39. Mohanty, A.K.; Senapati, M.R.; Beberta, S.; Lenka, S.K. Texture-based features for classification of mammograms using decision tree. Neural Comput. Appl. 2013, 23, 1011–1017.
  40. Liberini, V.; Laudicella, R.; Balma, M.; Nicolotti, D.G.; Buschiazzo, A.; Grimaldi, S.; Lorenzon, L.; Bianchi, A.; Peano, S.; Bartolotta, T.V.; et al. Radiomics and artificial intelligence in prostate cancer: New tools for molecular hybrid imaging and theragnostics. Eur. Radiol. Exp. 2022, 6, 27.
  41. Van Booven, D.J.; Kuchakulla, M.; Pai, R.; Frech, F.S.; Ramasahayam, R.; Reddy, P.; Parmar, M.; Ramasamy, R.; Arora, H. A Systematic Review of Artificial Intelligence in Prostate Cancer. Res. Rep. Urol. 2021, 13, 31–39.
  42. Wan, Y.L.; Wu, P.W.; Huang, P.C.; Tsay, P.K.; Pan, K.T.; Trang, N.N.; Chuang, W.Y.; Wu, C.Y.; Lo, S.B. The Use of Artificial Intelligence in the Differentiation of Malignant and Benign Lung Nodules on Computed Tomograms Proven by Surgical Pathology. Cancers 2020, 12, 2211.
  43. Heuvelmans, M.A.; van Ooijen, P.M.A.; Ather, S.; Silva, C.F.; Han, D.; Heussel, C.P.; Hickes, W.; Kauczor, H.-U.; Novotny, P.; Peschl, H.; et al. Lung cancer prediction by Deep Learning to identify benign lung nodules. Lung Cancer 2021, 154, 1–4.
  44. Li, X.; Hu, B.; Li, H.; You, B. Application of artificial intelligence in the diagnosis of multiple primary lung cancer. Thorac. Cancer 2019, 10, 2168–2174.
  45. Li, X.; Lu, Y.; Xiong, J.; Wang, D.; She, D.; Kuai, X.; Geng, D.; Yin, B. Presurgical differentiation between malignant haemangiopericytoma and angiomatous meningioma by a radiomics approach based on texture analysis. J. Neuroradiol. 2019, 46, 281–287.
  46. Ullah, N.; Khan, J.A.; Khan, M.S.; Khan, W.; Hassan, I.; Obayya, M.; Negm, N.; Salama, A.S. An Effective Approach to Detect and Identify Brain Tumors Using Transfer Learning. Appl. Sci. 2022, 12, 5645.
  47. Roodman, G.D. Skeletal imaging and management of bone disease. Hematol. Am. Soc. Hematol. Educ. Program 2008, 2008, 313–319.
  48. He, Y.; Pan, I.; Bao, B.; Halsey, K.; Chang, M.; Liu, H.; Peng, S.; Sebro, R.A.; Guan, J.; Yi, T.; et al. Deep learning-based classification of primary bone tumors on radiographs: A preliminary study. EBioMedicine 2020, 62, 103121.
  49. Umer, M.; Hasan, O.H.A.; Khan, D.; Uddin, N.; Noordin, S. Systematic approach to musculoskeletal benign tumors. Int. J. Surg. Oncol. 2017, 2, e46.
  50. Remotti, F.; Feldman, F. Nonneoplastic lesions that simulate primary tumors of bone. Arch. Pathol. Lab. Med. 2012, 136, 772–788.
  51. Teo, H.E.; Peh, W.C. Primary bone tumors of adulthood. Cancer Imaging 2004, 4, 74–83.
  52. Miller, T.T. Bone tumors and tumorlike conditions: Analysis with conventional radiography. Radiology 2008, 246, 662–674.
  53. Costelloe, C.M.; Madewell, J.E. Radiography in the initial diagnosis of primary bone tumors. AJR Am. J. Roentgenol. 2013, 200, 3–7.
  54. Gemescu, I.N.; Thierfelder, K.M.; Rehnitz, C.; Weber, M.-A. Imaging Features of Bone Tumors: Conventional Radiographs and MR Imaging Correlation. Magn. Reson. Imaging Clin. N. Am. 2019, 27, 753–767.
  55. Hayes, C.W.; Conway, W.F.; Sundaram, M. Misleading aggressive MR imaging appearance of some benign musculoskeletal lesions. Radiographics 1992, 12, 1119–1134; discussion 1135–1136.
  56. Camargo, O.P.; Croci, A.T.; Oliveira, C.R.; Baptista, A.M.; Caiero, M.T. Functional and radiographic evaluation of 214 aggressive benign bone lesions treated with curettage, cauterization, and cementation: 24 years of follow-up. Clinics 2005, 60, 439–444.
  57. Weber, M.A.; Papakonstantinou, O.; Nikodinovska, V.V.; Vanhoenacker, F.M. Ewing’s Sarcoma and Primary Osseous Lymphoma: Spectrum of Imaging Appearances. Semin. Musculoskelet Radiol. 2019, 23, 36–57.
  58. McCarville, M.B.; Chen, J.Y.; Coleman, J.L.; Li, Y.; Li, X.; Adderson, E.E.; Neel, M.D.; Gold, R.E.; Kaufman, R.A. Distinguishing Osteomyelitis From Ewing Sarcoma on Radiography and MRI. AJR Am. J. Roentgenol. 2015, 205, 640–650; quiz 651.
  59. Kuleta-Bosak, E.; Kluczewska, E.; Machnik-Broncel, J.; Madziara, W.; Ciupinska-Kajor, M.; Sojka, D.; Rogala, W.; Juszczyk, J.; Wilk, R. Suitability of imaging methods (X-ray, CT, MRI) in the diagnostics of Ewing’s sarcoma in children—Analysis of own material. Pol. J. Radiol. 2010, 75, 18–28.
  60. Mar, W.A.; Taljanovic, M.S.; Bagatell, R.; Graham, A.R.; Speer, D.P.; Hunter, T.B.; Rogers, L.F. Update on imaging and treatment of Ewing sarcoma family tumors: What the radiologist needs to know. J. Comput. Assist. Tomogr. 2008, 32, 108–118.
  61. Consalvo, S.; Hinterwimmer, F.; Neumann, J.; Steinborn, M.; Salzmann, M.; Seidl, F.; Lenze, U.; Knebel, C.; Rueckert, D.; Burgkart, R.H.H. Two-Phase Deep Learning Algorithm for Detection and Differentiation of Ewing Sarcoma and Acute Osteomyelitis in Paediatric Radiographs. Anticancer Res. 2022, 42, 4371–4380.
  62. Vanel, D.; Ruggieri, P.; Ferrari, S.; Picci, P.; Gambarotti, M.; Staals, E.; Alberghini, M. The incidental skeletal lesion: Ignore or explore? Cancer Imaging 2009, 9, S38–S43.
  63. Caballes, R.L.; Caballes, R.A., Jr. Polyostotic giant enostoses with strongly positive radionuclide bone scan. Ann. Diagn. Pathol. 2004, 8, 247–251.
  64. Hong, J.H.; Jung, J.Y.; Jo, A.; Nam, Y.; Pak, S.; Lee, S.Y.; Park, H.; Lee, S.E.; Kim, S. Development and Validation of a Radiomics Model for Differentiating Bone Islands and Osteoblastic Bone Metastases at Abdominal CT. Radiology 2021, 299, 626–632.
  65. Ulano, A.; Bredella, M.A.; Burke, P.; Chebib, I.; Simeone, F.J.; Huang, A.J.; Torriani, M.; Chang, C.Y. Distinguishing Untreated Osteoblastic Metastases From Enostoses Using CT Attenuation Measurements. Am. J. Roentgenol. 2016, 207, 362–368.
  66. Sala, F.; Dapoto, A.; Morzenti, C.; Firetto, M.C.; Valle, C.; Tomasoni, A.; Sironi, S. Bone islands incidentally detected on computed tomography: Frequency of enostosis and differentiation from untreated osteoblastic metastases based on CT attenuation value. Br. J. Radiol. 2019, 92, 20190249.
  67. Dong, Y.; Zheng, S.; Machida, H.; Wang, B.; Liu, A.; Liu, Y.; Zhang, X. Differential diagnosis of osteoblastic metastases from bone islands in patients with lung cancer by single-source dual-energy CT: Advantages of spectral CT imaging. Eur. J. Radiol. 2015, 84, 901–907.
  68. Schajowicz, F.; Sissons, H.A.; Sobin, L.H. The World Health Organization’s histologic classification of bone tumors. A commentary on the second edition. Cancer 1995, 75, 1208–1214.
  69. Clézardin, P.; Coleman, R.; Puppo, M.; Ottewell, P.; Bonnelye, E.; Paycha, F.; Confavreux, C.B.; Holen, I. Bone metastasis: Mechanisms, therapies, and biomarkers. Physiol. Rev. 2021, 101, 797–855.
  70. Li, Y.; Zhou, W.; Lv, G.; Luo, G.; Zhu, Y.; Liu, J. Classification of Bone Tumor on CT Images Using Deep Convolutional Neural Network. In Proceedings of the Artificial Neural Networks and Machine Learning—ICANN 2018, Rhodes, Greece, 4–7 October 2018; Lecture Notes in Computer Science. pp. 127–136.
  71. Cannavò, L.; Albano, D.; Messina, C.; Corazza, A.; Rapisarda, S.; Pozzi, G.; Di Bernardo, A.; Parafioriti, A.; Scotto, G.; Perrucchini, G.; et al. Accuracy of CT and MRI to assess resection margins in primary malignant bone tumours having histology as the reference standard. Clin. Radiol. 2019, 74, e713–e736.
  72. Nascimento, D.; Suchard, G.; Hatem, M.; de Abreu, A. The role of magnetic resonance imaging in the evaluation of bone tumours and tumour-like lesions. Insights Imaging 2014, 5, 419–440.
  73. James, S.L.; Panicek, D.M.; Davies, A.M. Bone marrow oedema associated with benign and malignant bone tumours. Eur. J. Radiol. 2008, 67, 11–21.
  74. Oh, E.; Yoon, Y.C.; Kim, J.H.; Kim, K. Multiparametric approach with diffusion-weighted imaging and dynamic contrast-enhanced MRI: A comparison study for differentiating between benign and malignant bone lesions in adults. Clin. Radiol. 2017, 72, 552–559.
  75. Hillengass, J.; Stieltjes, B.; Bäuerle, T.; McClanahan, F.; Heiss, C.; Hielscher, T.; Wagner-Gund, B.; Habetler, V.; Goldschmidt, H.; Schlemmer, H.P.; et al. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and diffusion-weighted imaging of bone marrow in healthy individuals. Acta Radiol. 2011, 52, 324–330.
  76. Suh, C.H.; Yun, S.J.; Jin, W.; Lee, S.H.; Park, S.Y.; Ryu, C.-W. ADC as a useful diagnostic tool for differentiating benign and malignant vertebral bone marrow lesions and compression fractures: A systematic review and meta-analysis. Eur. Radiol. 2018, 28, 2890–2902.
  77. Pozzi, G.; Albano, D.; Messina, C.; Angileri, S.A.; Al-Mnayyis, A.a.; Galbusera, F.; Luzzati, A.; Perrucchini, G.; Scotto, G.; Parafioriti, A.; et al. Solid bone tumors of the spine: Diagnostic performance of apparent diffusion coefficient measured using diffusion-weighted MRI using histology as a reference standard. J. Magn. Reson. Imaging 2018, 47, 1034–1042.
  78. Sharma, G.; Saran, S.; Saxena, S.; Goyal, T. Multiparametric evaluation of bone tumors utilising diffusion weighted imaging and dynamic contrast enhanced magnetic resonance imaging. J. Clin. Orthop. Trauma. 2022, 30, 101899.
  79. Kayhan, A.; Yang, C.; Soylu, F.N.; Lakadamyalı, H.; Sethi, I.; Karczmar, G.; Stadler, W.; Oto, A. Dynamic contrast-enhanced MR imaging findings of bone metastasis in patients with prostate cancer. World J. Radiol. 2011, 3, 241–245.
  80. Jung, H.-S.; Jee, W.-H.; McCauley, T.R.; Ha, K.-Y.; Choi, K.-H. Discrimination of Metastatic from Acute Osteoporotic Compression Spinal Fractures with MR Imaging1. RadioGraphics 2003, 23, 179–187.
  81. Board, W. Classification of Tumours Editorial. Soft Tissue and Bone Tumours; International Agency for Research on Cancer: Lyon, France, 2020.
  82. Van Praag Veroniek, V.M.; Rueten-Budde, A.J.; Ho, V.; Dijkstra, P.D.S.; Fiocco, M.; van de Sande, M.A.J. Incidence, outcomes and prognostic factors during 25 years of treatment of chondrosarcomas. Surg. Oncol. 2018, 27, 402–408.
  83. Skeletal Lesions Interobserver Correlation among Expert Diagnosticians (SLICED) Study Group. Reliability of Histopathologic and Radiologic Grading of Cartilaginous Neoplasms in Long Bones. J. Bone Jt. Surg. 2007, 89, 2113–2123.
  84. Eefting, D.; Schrage, Y.M.; Geirnaerdt, M.J.; Le Cessie, S.; Taminiau, A.H.; Bovée, J.V.; Hogendoorn, P.C. Assessment of interobserver variability and histologic parameters to improve reliability in classification and grading of central cartilaginous tumors. Am. J. Surg. Pathol. 2009, 33, 50–57.
  85. Verstraete, K.L.; Lang, P. Bone and soft tissue tumors: The role of contrast agents for MR imaging. Eur. J. Radiol. 2000, 34, 229–246.
  86. Kransdorf, M.J. The use of gadolinium in the MR evaluation of musculoskeletal tumors. Top. Magn. Reason. Imaging 1996, 8, 15–23.
  87. May, D.A.; Good, R.B.; Smith, D.K.; Parsons, T.W. MR imaging of musculoskeletal tumors and tumor mimickers with intravenous gadolinium: Experience with 242 patients. Skelet. Radiol. 1997, 26, 2–15.
  88. Sundaram, M. The use of gadolinium in the MR imaging of bone tumors. Semin Ultrasound CT MR 1997, 18, 307–311.
  89. Stacy, G.S.; Mahal, R.S.; Peabody, T.D. Staging of Bone Tumors: A Review with Illustrative Examples. Am. J. Roentgenol. 2006, 186, 967–976.
  90. Eweje, F.R.; Bao, B.; Wu, J.; Dalal, D.; Liao, W.H.; He, Y.; Luo, Y.; Lu, S.; Zhang, P.; Peng, X.; et al. Deep Learning for Classification of Bone Lesions on Routine MRI. EBioMedicine 2021, 68, 103402.
  91. Mithal, L.B.; Patel, P.S.; Mithal, D.; Palac, H.L.; Rozenfeld, M.N. Use of gadolinium-based magnetic resonance imaging contrast agents and awareness of brain gadolinium deposition among pediatric providers in North America. Pediatr. Radiol. 2017, 47, 657–664.
  92. Bhargava, R.; Hahn, G.; Hirsch, W.; Kim, M.J.; Mentzel, H.J.; Olsen, O.E.; Stokland, E.; Triulzi, F.; Vazquez, E. Contrast-enhanced magnetic resonance imaging in pediatric patients: Review and recommendations for current practice. Magn. Reason. Insights 2013, 6, 95–111.
  93. Strauss, L.G.; Conti, P.S. The applications of PET in clinical oncology. J. Nucl. Med. 1991, 32, 623–648; discussion 649–650.
  94. Hoh, C.K.; Schiepers, C.; Seltzer, M.A.; Gambhir, S.S.; Silverman, D.H.; Czernin, J.; Maddahi, J.; Phelps, M.E. PET in oncology: Will it replace the other modalities? Semin. Nucl. Med. 1997, 27, 94–106.
  95. Brock, C.S.; Meikle, S.R.; Price, P. Does fluorine-18 fluorodeoxyglucose metabolic imaging of tumours benefit oncology? Eur. J. Nucl. Med. 1997, 24, 691–705.
  96. Alauddin, M.M. Positron emission tomography (PET) imaging with (18)F-based radiotracers. Am. J. Nucl. Med. Mol. Imaging 2012, 2, 55–76.
  97. Weber, D.A.; Greenberg, E.J.; Dimich, A.; Kenny, P.J.; Rothschild, E.O.; Myers, W.P.; Laughlin, J.S. Kinetics of radionuclides used for bone studies. J. Nucl. Med. 1969, 10, 8–17.
  98. Blake, G.M.; Park-Holohan, S.J.; Cook, G.J.; Fogelman, I. Quantitative studies of bone with the use of 18F-fluoride and 99mTc-methylene diphosphonate. Semin. Nucl. Med. 2001, 31, 28–49.
  99. Schlonsky, J. Radioisotope scanning of bone. A review of the literature. Ohio. State Med. J. 1972, 68, 128–133.
  100. Bastawrous, S.; Bhargava, P.; Behnia, F.; Djang, D.S.W.; Haseley, D.R. Newer PET Application with an Old Tracer: Role of 18F-NaF Skeletal PET/CT in Oncologic Practice. RadioGraphics 2014, 34, 1295–1316.
  101. Kern, K.A.; Brunetti, A.; Norton, J.A.; Chang, A.E.; Malawer, M.; Lack, E.; Finn, R.D.; Rosenberg, S.A.; Larson, S.M. Metabolic imaging of human extremity musculoskeletal tumors by PET. J. Nucl. Med. 1988, 29, 181–186.
  102. Adler, L.P.; Blair, H.F.; Makley, J.T.; Williams, R.P.; Joyce, M.J.; Leisure, G.; al-Kaisi, N.; Miraldi, F. Noninvasive grading of musculoskeletal tumors using PET. J. Nucl. Med. 1991, 32, 1508–1512.
  103. Griffeth, L.K.; Dehdashti, F.; McGuire, A.H.; McGuire, D.J.; Perry, D.J.; Moerlein, S.M.; Siegel, B.A. PET evaluation of soft-tissue masses with fluorine-18 fluoro-2-deoxy-D-glucose. Radiology 1992, 182, 185–194.
  104. Fan, X.; Zhang, X.; Zhang, Z.; Jiang, Y. Deep Learning-Based Identification of Spinal Metastasis in Lung Cancer Using Spectral CT Images. Sci. Program. 2021, 2021, 2779390.
  105. Orlhac, F.; Thézé, B.; Soussan, M.; Boisgard, R.; Buvat, I. Multiscale Texture Analysis: From 18F-FDG PET Images to Histologic Images. J. Nucl. Med. 2016, 57, 1823–1828.
  106. Orlhac, F.; Soussan, M.; Maisonobe, J.-A.; Garcia, C.A.; Vanderlinden, B.; Buvat, I. Tumor Texture Analysis in 18F-FDG PET: Relationships between Texture Parameters, Histogram Indices, Standardized Uptake Values, Metabolic Volumes, and Total Lesion Glycolysis. J. Nucl. Med. 2014, 55, 414.
  107. Stacy, G.S.; Dixon, L.B. Pitfalls in MR image interpretation prompting referrals to an orthopedic oncology clinic. Radiographics 2007, 27, 805–826.
  108. Trieu, J.; Sinnathamby, M.; Di Bella, C.; Pianta, M.; Perera, W.; Slavin, J.L.; Schlicht, S.M.; Choong, P.F. Biopsy and the diagnostic evaluation of musculoskeletal tumours: Critical but often missed in the 21st century. ANZ J. Surg. 2016, 86, 133–138.
  109. Meek, R.D.; Mills, M.K.; Hanrahan, C.J.; Beckett, B.R.; Leake, R.L.; Allen, H.; Williams, D.D.; Tommack, M.; Schmahmann, S.; Hansford, B.G. Pearls and Pitfalls for Soft-Tissue and Bone Biopsies: A Cross-Institutional Review. RadioGraphics 2020, 40, 266–290.
  110. Traina, F.; Errani, C.; Toscano, A.; Pungetti, C.; Fabbri, D.; Mazzotti, A.; Donati, D.; Faldini, C. Current concepts in the biopsy of musculoskeletal tumors: AAOS exhibit selection. J. Bone Jt. Surg. Am. 2015, 97, e7.
  111. Tan, A.E.; Ng, D.C. Differentiating osteoradionecrosis from nasopharyngeal carcinoma tumour recurrence using 99Tcm-sestamibi SPECT/CT. Br. J. Radiol. 2011, 84, e172–e175.
  112. Alhilali, L.; Reynolds, A.R.; Fakhran, S. Osteoradionecrosis after radiation therapy for head and neck cancer: Differentiation from recurrent disease with CT and PET/CT imaging. AJNR Am. J. Neuroradiol. 2014, 35, 1405–1411.
  113. Lipkova, J.; Chen, R.J.; Chen, B.; Lu, M.Y.; Barbieri, M.; Shao, D.; Vaidya, A.J.; Chen, C.; Zhuang, L.; Williamson, D.F.K.; et al. Artificial intelligence for multimodal data integration in oncology. Cancer Cell 2022, 40, 1095–1110.
  114. Antropova, N.; Huynh, B.Q.; Giger, M.L. A deep feature fusion methodology for breast cancer diagnosis demonstrated on three imaging modality datasets. Med. Phys. 2017, 44, 5162–5171.
More
© Text is available under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/)
Academic Video Service
Feedback