CT Radiomics Helps Classify Small Lung Nodules
By MedImaging International staff writers Posted on 01 Feb 2021 |
Image: CT radiomics can help classify lung nodule malignancy (Photo courtesy of Getty Images)
A machine-learning (ML) algorithm can be highly accurate for classifying very small lung nodules found in low-dose CT lung screening programs, according to a new study.
Researchers at the BC Cancer Research Center (BCCRC; Vancouver, Canada) trained a linear discriminant analysis (LDA) ML algorithm--using data from the Pan-Canadian Early Detection of Lung Cancer (PanCan) study--to characterize, analyze, and classify small lung nodules as malignant or benign by extracting approximately 170 texture and shape radiomic features, following semi-automated nodule segmentation on the images. They then compared the performance of the algorithm with that of the Prostate, Lung, Colorectal, and Ovarian (PLCO) m2012 malignancy risk score calculator on another dataset.
The study cohort consisted of 139 malignant nodules and 472 benign nodules that were approximately matched in size. The researchers applied size restrictions (based on Lung-RADS classification criteria) to remove any nodules from the dataset that would already be considered suspicious, which would include any nodule with solid components greater than 8 mm in diameter. The results showed the ML algorithm significantly outperformed the (PLCO) m2012 risk-prediction model, especially when demographic data were added to radiomics analysis. The study was presented at the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging, held during January 2021.
“The best results were achieved in a subset of patients who were younger than 64, female, did not have emphysema, smoked fewer than 42 pack years, did not have a family history of lung cancer, and were not current smokers,” said senior author and study presenter Rohan Abraham, PhD. “Combined with clinician expertise and experience, this has the potential to enable earlier intervention and reduce the need for follow-up CT.”
Current lung nodule classification relies on nodule size, a factor that is of limited use for sub-centimeter nodules, or on volume doubling time, a variable that requires follow-up CT exams. As a result, very small lung nodules, with solid components of less than 8 mm in diameter (and therefore below the Lung-RADS 4A risk-stratification threshold), are very difficult to classify, and they are often given a "wait and see" management plan.
Related Links:
BC Cancer Research Center
Researchers at the BC Cancer Research Center (BCCRC; Vancouver, Canada) trained a linear discriminant analysis (LDA) ML algorithm--using data from the Pan-Canadian Early Detection of Lung Cancer (PanCan) study--to characterize, analyze, and classify small lung nodules as malignant or benign by extracting approximately 170 texture and shape radiomic features, following semi-automated nodule segmentation on the images. They then compared the performance of the algorithm with that of the Prostate, Lung, Colorectal, and Ovarian (PLCO) m2012 malignancy risk score calculator on another dataset.
The study cohort consisted of 139 malignant nodules and 472 benign nodules that were approximately matched in size. The researchers applied size restrictions (based on Lung-RADS classification criteria) to remove any nodules from the dataset that would already be considered suspicious, which would include any nodule with solid components greater than 8 mm in diameter. The results showed the ML algorithm significantly outperformed the (PLCO) m2012 risk-prediction model, especially when demographic data were added to radiomics analysis. The study was presented at the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging, held during January 2021.
“The best results were achieved in a subset of patients who were younger than 64, female, did not have emphysema, smoked fewer than 42 pack years, did not have a family history of lung cancer, and were not current smokers,” said senior author and study presenter Rohan Abraham, PhD. “Combined with clinician expertise and experience, this has the potential to enable earlier intervention and reduce the need for follow-up CT.”
Current lung nodule classification relies on nodule size, a factor that is of limited use for sub-centimeter nodules, or on volume doubling time, a variable that requires follow-up CT exams. As a result, very small lung nodules, with solid components of less than 8 mm in diameter (and therefore below the Lung-RADS 4A risk-stratification threshold), are very difficult to classify, and they are often given a "wait and see" management plan.
Related Links:
BC Cancer Research Center
Latest Radiography News
- Novel Breast Imaging System Proves As Effective As Mammography
- AI Assistance Improves Breast-Cancer Screening by Reducing False Positives
- AI Could Boost Clinical Adoption of Chest DDR
- 3D Mammography Almost Halves Breast Cancer Incidence between Two Screening Tests
- AI Model Predicts 5-Year Breast Cancer Risk from Mammograms
- Deep Learning Framework Detects Fractures in X-Ray Images With 99% Accuracy
- Direct AI-Based Medical X-Ray Imaging System a Paradigm-Shift from Conventional DR and CT
- Chest X-Ray AI Solution Automatically Identifies, Categorizes and Highlights Suspicious Areas
- AI Diagnoses Wrist Fractures As Well As Radiologists
- Annual Mammography Beginning At 40 Cuts Breast Cancer Mortality By 42%
- 3D Human GPS Powered By Light Paves Way for Radiation-Free Minimally-Invasive Surgery
- Novel AI Technology to Revolutionize Cancer Detection in Dense Breasts
- AI Solution Provides Radiologists with 'Second Pair' Of Eyes to Detect Breast Cancers
- AI Helps General Radiologists Achieve Specialist-Level Performance in Interpreting Mammograms
- Novel Imaging Technique Could Transform Breast Cancer Detection
- Computer Program Combines AI and Heat-Imaging Technology for Early Breast Cancer Detection