Application of the Machine Learning Model in Classification of Hepatic Lesions Based on Time-Signal Intensity Curve on Triphasic Contrast Enhanced MRI and Role of MR Radiomics in Diagnosis of Hepatocellular Carcinoma
试验速览
- 阶段
- 不适用
- 状态
- 尚未招募
- 发起方
- NA
- 入组人数
- 120
- 试验地点
- 1
- 主要终点
- The machine learning or deep learning model based on a time signal intensity curve can be used for better classification of hepatic lesions than visual assessment.
研究概览
简要总结
Hepatic lesions are commonly detected on imaging and frequently present with diagnostic dilemmas. The diagnosis of liver malignancy by alpha-fetoprotein (AFP) was 67.8-74.4%. Though biopsy and histopathological confirmation are required to diagnose the hepatic lesion because of their invasive nature, these methods are not ideal and need to be or are usually supported by various imaging modalities
In MRI conventionally, the characterisation of the hepatic lesions in dynamic post-contrast MRI depends on visual observation of the enhancement pattern of the lesion, which reflects the physiological process. A recently developed tracer kinetic model using a time-signal intensity curve helps in the quantitative analysis of the enhancement pattern of the lesion in different phases.
Tumour Segmentation for time signal intensity curve
The ROI will be drawn manually on hepatic tumors in pre-contrast T1W, arterial, venous and delay phases of triphasic contrast phases for hepatic lesions.
Extraction of Features
3D slicer (version 5.1.0) will be used to identify the mean value of hepatic lesion and aorta.
Machine Learning Model
For the time-signal intensity curve, a machine learning or deep learning model will be developed using MATLAB or PYTHON platform and trained and tested using a retrospective data set.
The obtained curve will be divided into: i)increase rapidly and decrease rapidly, ii)increase rapidly and decrease slowly, iii)increase slowly and decrease slowly, iv) increase slowly after no apparent decline. Maximum slope of increase (MSI) and maximum slope of decrease (MSD) values will be identified.
Validation of Model
Region of interest (ROI) will be drawn manually in the arterial phase and same ROI will be used in the series of venous and delay phase images. Using developed model, the time signal intensity curve will be obtained for hepatic lesions.
The obtained curve will be divided into: i)increase rapidly and decrease rapidly, ii)increase rapidly and decrease slowly, iii)increase slowly and decrease slowly, iv) increase slowly after no apparent decline. Maximum slope of increase (MSI) and maximum slope of decrease (MSD) values will be identified.
Tumour Segmentation for radiomics features in HCC
The ROI will be drawn manually on hepatic tumors in DWI, pre-contrast T1W, arterial, venous and delayed phase of triphasic contrast images using a 3D slicer.
Extraction of Features
3D slicer (version 5.1.0) will be used for the extraction of radiomics features. Any variation in features intensity normalization will be done.
Machine Learning Model
Machine learning model will be developed using MATLAB or PYTHON platform and trained and tested using a retrospective data set.
Radiomic Feature Extraction
The DWI and triphasic contrast images will be uploaded in a 3D slicer. The ROI will be drawn manually on the hepatic tumor in DWI, pre-contrast T1W image, arterial, venous, and delay phase of triphasic contrast phases and radiomics features will be extracted. Any variation in features intensity normalization will be done.
Validation of Model
Using the developed models, HCC and Non-HCC will be classified.
研究设计
- 研究类型
- Observational
入排标准
- 年龄范围
- 18.00 Year(s) 至 80.00 Year(s)（—）
- 性别
- All
入选标准
- •Study with triphasic contrast MRI of Abdomen showing hepatic lesions.
- •Age group of 18- 80 years.
排除标准
- •Patients with a history of trauma Patients with a history of surgery/ radiotherapy for hepatic lesions MRI Abdomen without contrast.
- •Artefacts present in the area of interest.
结局指标
主要结局
The machine learning or deep learning model based on a time signal intensity curve can be used for better classification of hepatic lesions than visual assessment.
时间窗: The machine learning or deep learning model based on a time signal intensity curve can be used for better classification of hepatic lesions than visual assessment. | • The machine learning based on MR radiomics features can be used to improve the diagnosis of HCC.
• The machine learning based on MR radiomics features can be used to improve the diagnosis of HCC.
时间窗: The machine learning or deep learning model based on a time signal intensity curve can be used for better classification of hepatic lesions than visual assessment. | • The machine learning based on MR radiomics features can be used to improve the diagnosis of HCC.
次要结局
未报告次要终点