Risk Stratification of Orbital Tumors Based on MRl and Artificial Intelligence

Orbital tumors can be categorized into benign and malignant tumors, and there are significant variations in their biological behavior, treatment, and prognosis. This study aims to enhance the accurate diagnosis and risk stratification of orbital tumors using artificial intelligence (AI) technology and multiparameter magnetic resonance imaging (MRI) data. It further explores the intrinsic relationship between MRI and the differential diagnosis of benign and malignant orbital tumors, as well as the pathological subtypes of malignant tumors and Ki-67 expression levels. This research aims to aid in guiding personalized diagnosis and treatment decision-making for patients with orbital tumors while promoting the practical application and incorporation of AI technology.

Study Overview

• Status: Completed
• Conditions: Orbital Neoplasms
• Intervention / Treatment: Other: Multi-parametric MRI and image analysis by deep learning or machine learning algorithms

Detailed Description

Although orbital tumors are less common than other eye-related diseases, they can be extremely detrimental to patients. Not only can they cause physical disfigurement, but they can also lead to functional impairments such as diminished vision and restricted eye movement. Orbital tumors can be categorized as either benign or malignant, and there are significant disparities in their biological behavior, treatment approaches, outcomes, and prognosis, which complicates the processes of differential diagnosis and treatment selection. For malignant lesions, the treatment plans and prognosis of patients vary due to the different pathological types and stages. Hence, there is a pressing clinical necessity to devise accurate diagnostic methods for orbital tumors. Multiparametric magnetic resonance imaging (mp-MRI) currently stands as the leading non-invasive imaging technique for diagnosing orbital tumors. This study is centered on precise diagnosis of orbital tumor risk stratification, utilizing artificial intelligence algorithm technology to explore the inherent connection between MRI images and the distinguishing diagnosis of benign and malignant orbital tumors, histological types and Ki-67 expression levels of malignant tumors. It aims to integrate clinical information and quantitative MRI features to construct prediction models, aid in guiding individual diagnosis and treatment decisions for patients with orbital tumors and facilitate the application and advancement of artificial intelligence technology. Specifically, the research objectives are outlined as follows:

1. Establishing a deep learning-based automatic segmentation model for orbital tumors using a multi-sequence MRI dataset from multiple centers, thereby reducing the time required for manual delineation and proving beneficial for subsequent analysis.
2. Developing a model for identifying malignant and benign orbital tumors using multiple machine learning algorithms combined with multi-sequence MRI dataset, with the aim of providing more precise information for distinguishing between these two entities.
3. Constructing robust diagnostic models using machine learning or deep learning approaches with quantitative multi-sequence MRI features to identify the histological type and Ki-67 expression levels of malignant orbital tumors, with the purpose of enhancing detection rates and accuracy, thereby achieving risk stratification for patients with malignant orbital tumors.

• Study Type: Observational
• Enrollment (Actual): 600

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Patients diagnosed with malignant or benign orbital tumors confirmed by pathology, who underwent multiparametric MRl (mp-MRl) at BeiiingTongren Hospital from 2015 to 2022, were included in this research. Otherwise, patients lacking a definitive pathological diagnosis or pre-operative multiparametric MRl (mp-MRl) were excluded from this investigation.

Description

Inclusion Criteria:

• The patients with orbital tumors who underwent pre-operative multiparametricMRl (mp-MRl) at Beijing Tongren Hospital from 2015 to 2022.

Exclusion Criteria:

• The patients without pre-operative multiparametric MRl (mp-MRl) or clear pathological diagnosis.

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Malignant orbital tumors; Patients with malignant orbital tumors (lymphoma, melanoma, ...) diagnosed by pathological confirmation.	Other: Multi-parametric MRI and image analysis by deep learning or machine learning algorithms; Diagnosis models are established using quantitative features extracted from the multi-parametric MRI images and further processed by appropriate deep learning or machine learning algorithms.
Benign orbital tumors; Patients with benign orbital tumors (cavernous hemangioma, inflammatory pseudotumor, ...) diagnosed by pathological confirmation.	Other: Multi-parametric MRI and image analysis by deep learning or machine learning algorithms; Diagnosis models are established using quantitative features extracted from the multi-parametric MRI images and further processed by appropriate deep learning or machine learning algorithms.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
The area under the curve of Receiver Operating Characteristic of the diagnostic models for the differential diagnosis of malignant and benign orbital tumors, high and low grades of histological types, and levels of Ki-67 expression in malignant ones.	The area under the ROC curve is calculated by integrating the ROC curve, which plots Sensitivity against 1 - Specificity.	Pre-operation

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
The area under the Precision-Recall curve of the diagnostic models for the differential diagnosis of malignant and benign orbital tumors, high and low grades of histological types, and high and low levels of Ki-67 expression in malignant orbital tumors.	The area under the precision-recall curve is determined by integrating the Precision-Recall curve, which plots Precision against Recall.	Pre-operation
Sensitivity of the diagnostic models for the differential diagnosis of malignant and benign orbital tumors, high and low grades of histological types, and high and low levels of Ki-67 expression in malignant orbital tumors.	Sensitivity is calculated as the ratio of true positives to the sum of true positives and false negatives.	Pre-operation
Specificity of the diagnostic models for the differential diagnosis of malignant and benign orbital tumors, high and low grades of histological types, and high and low levels of Ki-67 expression in malignant orbital tumors.	Specificity is calculated as the ratio of true negatives to the sum of true negatives and false positives.	Pre-operation
Accuracy of the diagnostic models for the differential diagnosis of malignant and benign orbital tumors, high and low grades of histological types, and high and low levels of Ki-67 expression in malignant orbital tumors.	Accuracy is calculated as the ratio of the sum of true positives and true negatives to the total number of cases.	Pre-operation

Collaborators and Investigators

• Sponsor: Beijing Tongren Hospital
• Investigators: Study Chair: Junfang Xian, M.D., Ph.D., Department of Radiology, Beijing Tongren Hospital, Capital Medical University

Study record dates

Study Major Dates

• Study Start (Actual): January 1, 2012
• Primary Completion (Actual): October 31, 2022
• Study Completion (Actual): December 31, 2023

Study Registration Dates

• First Submitted: March 22, 2024
• First Submitted That Met QC Criteria: March 22, 2024
• First Posted (Actual): March 28, 2024

Study Record Updates

• Last Update Posted (Actual): March 29, 2024
• Last Update Submitted That Met QC Criteria: March 28, 2024
• Last Verified: March 1, 2024

More Information

Terms related to this study

• Keywords: AI; Multiparametric MRI; Orbital Neoplasms
• Additional Relevant MeSH Terms: Neoplasms; Neoplasms by Site; Eye Diseases; Musculoskeletal Diseases; Bone Diseases; Orbital Diseases; Eye Neoplasms; Skull Neoplasms; Bone Neoplasms; Orbital Neoplasms
• Other Study ID Numbers: TREC2023-KY107

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No