OBB-0011
Section 1 - Basic information about you and your application:
Title of research project
Advancing FAI Surgery Outcomes through AI-based Radiomics Analysis of Hip MRI
Project summary
Femoroacetabular impingement (FAI) poses a significant challenge in orthopaedics, often necessitating surgical intervention. However, a substantial proportion of patients do not experience the expected benefits post-FAI surgery. Current clinical assessments for FAI are hindered by symptom non-specificity and limited radiographic reliability, exacerbating the need for a more sophisticated preoperative approach. Radiomics, an emerging field in medical imaging, offers a promising avenue to extract nuanced information from images, surpassing human capabilities. Building upon a robust radiomic foundation, this collaborative study between Imperial College and The University of Oxford will apply advanced techniques to hip MRI data collected from the FAIT trial. Machine learning algorithms will then evaluate correlations between Patient-Reported Outcome Measures (PROMs) and radiomic features, providing valuable insights into the predictive potential of radiomics in evaluating clinical outcomes post-FAI surgery. This project represents a pioneering step in enhancing FAI surgery outcomes through cutting-edge technology and multidisciplinary collaboration.
Type of project
Research
Type of research
Clinical
Specialty/Area:
Machine Learning, Radiomics, Femoroacetabular impingement
Start date
03/06/2024
End date
28/07/2023
Section 2 - Purpose of the research and originality
Aims / Objectives:
The aims of this study are to: 1) develop AI-based Radiomics analysis of hip MRI for the assessment of patients with FAIS, and 2) evaluate the association between radiomics features of FAI and patient-reported outcomes of FAIS patients who underwent hip arthroscopic surgery.
Background to investigation:
Femoroacetabular impingement (FAI) describes the pathological abutment of the femur on the acetabulum, either due to abnormal femoral head-neck morphology (cam-type FAI), acetabular morphology (pincer-type FAI), or a ‘mixed’ aetiology. FAI results in reduced range of motion and pain in deep flexion and internal rotation, and is often associated with damage to the labrum and the chondrolabral junction. Symptomatic FAI typically presents in active, young adults (1).
The magnitude of FAI pathomorphology can be measured on plain radiographs, CT imaging, and MR imaging (MRI). In 2D, the alpha-angle quantifies cam-type femoral morphology, and the lateral centre edge angle (LCEA) quantifies pincer-type acetabular over-coverage. 3D analysis of CT has enabled measurement of the alpha angle at multiple points around the femoral head-neck junction (Clinical Graphics, Den Haag, The Netherlands). However, these have highly variable sensitivity and specificity for diagnosing FAI, as 2D measurements of 3D anatomy are affected by patient position (2).
The combination of imaging findings correlated with this clinical presentation is termed FAI Syndrome (FAIS) (3). Two seminal randomised controlled trials have demonstrated that repairing the labrum, and reshaping the femoral head-neck junction and/or acetabular rim through hip arthroscopy is more effective than conservative management.(5,6) However, approximately one third of patients undergoing FAI surgery do not gain a clinically meaningful benefit. Those with advanced arthritis are less likely to improve with surgery, but generally the negative predictors are less well understood (6). Plain radiographic assessments have demonstrated limited reliability for differentiating between a healthy hip joint and early osteoarthritis (OA); the magnitude of the cam- or pincer-morphology is also not predictive of outcome (7).
MRI is a reference modality for identifying labral tears, subchondral and paralabral cysts, and subtle osteophytes in patients with FAI undergoing arthroscopy. MRI changes of advanced arthritis are correlated surgical outcomes (8). However, these macroscopic anatomical findings in MRI have poor specificity and negative predictive value (9), and the utility of MRI in patients with early/no OA is unproven.
This underscores the urgent need for a more sophisticated approach to preoperative assessment. Radiomics, a burgeoning field in medical imaging, offers a promising avenue for obtaining nuanced information from images that surpasses the ability of a human making direct measurements. A multitude of imaging features are extracted from a region of interest and subsequently reduced and selected to convey diagnostic or prognostic information. The assumption of radiomics is that image features quantify crucial information regarding pathologic conditions through intra-region heterogeneity. For the hip, radiomics takes 190 measurements; these are grouped into (i) intensity and histogram based first order statistics (FOS) features, (ii) texture features, and (iii) shape and size features. Deep radiomics uses convolutional neural networks (CNNs, a type of artificial intelligence (AI)) to directly extract features and obviate the need for predefined features. MRI-based radiomics has recently been validated for diagnosing FAI more accurately and rapidly than conventional measurement techniques (2). However, these have focused on shape and signal characteristics of the femoral and acetabular bone contour alone, without analyses of chondral surfaces, the labrum, or subchondral oedema – relevant to the clinical presentation of FAIS.
The proposed study will develop AI-Radiomics to interrogate MRI for shape, and chondral, labral and subchondral features related to FAIS. It will develop AI-based techniques, particularly those employing shape and gradient analyses to extract radiomic features from MRI. Finally, it will validate if these features are predictive of patient-reported outcomes after hip arthroscopy – thus assessing applicability of AI-Radiomics to clinical practice.
References:
- Palmer A, Fernquest S, Gimpel M, Birchall R, Judge A, Broomfield J, et al. Physical activity during adolescence and the development of cam morphology: a cross-sectional cohort study of 210 individuals. British Journal of Sports Medicine. 2018 May 1;52(9):601–10.
- Montin E, Kijowski R, Youm T, Lattanzi R. A radiomics approach to the diagnosis of femoroacetabular impingement. Frontiers in Radiology.2023 Mar 20;3:1151258.
- Griffin DR, Dickenson EJ, O’Donnell J, Agricola R, Awan T, Beck M, et al. The Warwick Agreement on Femoroacetabular Impingement Syndrome (FAI syndrome): an International Consensus Statement. British Journal of Sports Medicine. 2016 Sep 14;50(19):1169–76.
- Palmer AJR, Ayyar Gupta V, Fernquest S, Rombach I, Dutton SJ, Mansour R, et al. Arthroscopic hip surgery compared with physiotherapy and activity modification for the treatment of symptomatic femoroacetabular impingement: multicentre randomised controlled trial. BMJ. 2019 Feb 7;185.
- Griffin DR, Dickenson EJ, Wall PDH, Achana F, Donovan JL, Griffin J, et al. Hip arthroscopy versus best conservative care for the treatment of femoroacetabular impingement syndrome (UK FASHIoN): a multicentre randomised controlled trial. The Lancet. 2018 Jun;391(10136):2225–35.
- Andronic O, Claydon-Mueller LS, Cubberley R, Karczewski D, Sunil-Kumar KH, Khanduja V. Inconclusive and Contradictory Evidence for Outcomes After Hip Arthroscopy in Patients With Femoroacetabular Impingement and Osteoarthritis of Tönnis Grade 2 or Greater: A Systematic Review. Arthroscopy: The Journal of Arthroscopic & Related Surgery. 2022 Jul;38(7):2307-2318.e1.
- Jacobs CA, Burnham JM, Jochimsen KN, Molina D, Hamilton DA, Duncan ST. Preoperative Symptoms in Femoroacetabular Impingement Patients Are More Related to Mental Health Scores Than the Severity of Labral Tear or Magnitude of Bony Deformity. The Journal of Arthroplasty. 2017 Dec;32(12):3603–6.
- Conaway W, Agrawal R, Skelley NW, Waryasz GR, Small KM, Shah N, et al. MRA Findings Predictive of Hip Arthroscopy Outcomes for Femoroacetabular Impingement. Arthroscopy: The Journal of Arthroscopic & Related Surgery. 2018 Dec 1;34(12):e17–8.
- Annabell L, Master V, Rhodes A, Moreira B, Coetzee C, Tran P. Hip pathology: the diagnostic accuracy of magnetic resonance imaging. Journal of Orthopaedic Surgery and Research. 2018 May 29;13(1).
Section 3 - Plan of investigation
Plan of investigation:
Patients and imaging:
In this ethically approved study, 112 patients who have undergone hip arthroscopy during the multicentre Femoroacetabular Impingement Trial (FAIT) RCT will be included (1). We will analyse a complete dataset of patients with FAIS, including plain radiographs, MRIs with FAI-specific protocol, and systematically collected pre- and post-operative Patient-Reported Outcome Measures (PROMs) data.
Radiomics development:
We will use open-source software to systematically extract a previously validated array of quantitative features for hip shape (2), and develop quantitative features for chondral and labral pathology (3) from the dedicated hip MRI scans within the FAIT dataset. This process will involve the precise characterization of various radiomic parameters related to shape, texture, and signal intensity of FAIS – which will be entirely novel. A previous study has validated radiomics for FAI morphology using 17 scans, so this study will be adequately powered.(2)
Machine Learning Correlation Analysis:
The augmented dataset, labelled with radiomic features, will serve as the basis for our correlation analysis. Machine learning techniques will systematically investigate the relationship between changes in pre- and post-operative PROMs, and the radiomic features.
Validation and Quality Control:
Throughout the analysis, we will implement rigorous validation and quality control measures to ensure the accuracy, reliability, and reproducibility of our findings. This will involve cross-validation techniques, sensitivity analyses, and independent validation on subsets of the data (separating training, test, and validation data for ML).
Ethical Considerations:
This study will strictly adhere to all ethical guidelines and regulations governing the use of patient data, in line with the FAIT RCT data (4). All data will be anonymized and stored on secure university servers.
Expenses:
The funding will purchase software licences and computational cloud-based machine-learning server time necessary for radiomics analysis, data augmentation, and model-training.
References:
- Palmer AJR, Ayyar Gupta V, Fernquest S, Rombach I, Dutton SJ, Mansour R, et al. Arthroscopic hip surgery compared with physiotherapy and activity modification for the treatment of symptomatic femoroacetabular impingement: multicentre randomised controlled trial. BMJ. 2019 Feb 7;185.
- Montin E, Kijowski R, Youm T, Lattanzi R. A radiomics approach to the diagnosis of femoroacetabular impingement. Frontiers in Radiology.2023 Mar 20;3:1151258.
- Zhen T, Fang J, Hu D, Ruan M, Wang L, Fan S, et al. Risk stratification by nomogram of deep learning radiomics based on multiparametric magnetic resonance imaging in knee meniscus injury. International Orthopaedics. 2023 Oct 1;47(10):2497–505.
- Palmer AJR, Ayyar-Gupta V, Dutton SJ, Rombach I, Cooper CD, Pollard TC, et al. Protocol for the Femoroacetabular Impingement Trial (FAIT). Bone & Joint Research. 2014 Nov;3(11):321–7.
Section 4 - Research impact and benefits
Research impact and benefits:
Engaging in this project presents a tremendous opportunity to elevate my research acumen and tackle intricate clinical dilemmas in the realm of hip surgery. By immersing myself in this project, I aim to cultivate a profound expertise in machine learning, with a specific focus on its application in advanced imaging techniques. Furthermore, this project will provide me invaluable exposure to multicentre and multidisciplinary collaborations. This project stands as a stepping stone in my academic journey, bestowing upon me the essential knowledge, skills, and insights to forge lasting contributions in the field of hip surgery. It promises not only to my proficiency in research methodologies but also to provide me with a comprehensive understanding of the interactions between the disciplines of computational biology, imaging, and orthopaedic surgery.
Section 5 - Strength of individual and achievements to aid facilitate the project?
Name:
Mr. Hariharan Subbiah Ponniah
Institution:
MSk Lab / Imperial College London
Year of study / grade / job title:
Year 6
Graduation date:
28/06/2024
Are you undertaking this project as part of an intercalated degree?
No
Does your institution offer an intercalated degree option?
Yes
Role
My role involves working with data-scientists to lead the radiomics analysis on hip MRI data from the FAIT trial. This includes extracting radiomic features, employing data augmentation, and utilising machine learning algorithms. This builds on my existing experience in developing machine learning for hip surgery. (1) 1. Hariharan Subbiah Ponniah, Thomas Edwards, Jonathan Lex, Ross Davidson, Mustafa Al-Zubaidy, Irrum Afzal, Richard Field, Alexander Liddle, Justin Cobb, Kartik Logishetty. Machine learning can predict difficulty in anterior approach total hip arthroplasty, to improve patient safety and surgical training. European Orthopaedic Research Society 2023
Supervisor name
Kartik Logishetty
Supervisor job title
NIHR Clinical Lecturer in Trauma & Orthopaedics
Supervisor address
MSk Lab, Imperial College London 2nd Floor, Sir Michael Uren Hub 86 Wood Lane, London, London W12 0BZ, GB
Applicant reference
Please refer to the attached reference letter
Other documents
Applicant CV: Download hereReference letter: Download here
Confirmation of support: Download here