Abstract:e13639 Background: Automated Machine Learning (Auto-ML) in medical imaging is a process that allows non-experts to utilize machine learning techniques, opening the door for non-coder physician-driven exploitation of the technology. Auto-ML was applied for opportunistic detection of thyroid nodules in the context of low-dose lung cancer screening chest CT, facilitated by an innovative platform integration. By leveraging scans originally intended for lung cancer screening, suspicious appearing asymptomatic thyroid nodules can also be screened for where technically feasible. Methods: CT scans from the National Lung Screening Trial (NLST) dataset were utilized. This dataset contains low-dose chest CT examinations initially used for lung cancer screening. Annotation was carried out within a no-code web-based platform, Gesund.ai. A board-certified diagnostic radiologist annotated 100 CT examinations. A case was labeled "suspicious" if it presented a thyroid nodule of 1 cm or greater in diameter, while cases with no such nodules were deemed "not suspicious." The Auto-ML feature was then engaged within the software platform for parameter adjustments to refine the model without the need for manual coding. This phase was characterized by an iterative exploration of model performance, facilitated by the platform’s robust validation tools. Results: Upon training and validation, the Auto-ML model, cultivated from 100 annotated cases and assessed against a separate set of 130 cases, demonstrated a discernment accuracy of 0.51 in identifying "suspicious" thyroid nodules with a sensitivity and precision of 0.51 and 0.62 respectively. The AUC for this class stands at 0.69, indicating a moderate ability to distinguish between the presence and absence of thyroid nodules. This metric, while foundational, offers insight into the model's potential efficacy in real-world diagnostic scenarios, reflecting the initial capabilities of the platform's Auto-ML integration in enhancing diagnostic processes. Conclusions: The findings from this study highlight the potential of Auto-ML in revolutionizing opportunistic screening for thyroid nodules via lung cancer CT. The initial accuracy rate underscores the necessity for further refinement and validation of the model, but also demonstrates the promising potential of Auto-ML. Institutional Auto-ML would include extraction of appropriate studies through the no-code software platform integrated directly within the PACS framework. This process would ensure data immobility and foster a secure environment for dataset creation. Annotation would be carried out by multiple physician imagers within the same facility, enhancing the data’s accuracy and comprehensiveness specific for the institution’s patient population. This exploration highlights the strategic role of Auto-ML in advancing cancer patient care through innovative technological integration.

Automated Interstitial Lung Abnormality Probability Prediction at CT: A Stepwise Machine Learning Approach in the Boston Lung Cancer Study

Interstitial lung disease diagnosis and prognosis using an AI system integrating longitudinal data

Interstitial lung abnormalities (ILA) on routine chest CT: Comparison of radiologists' visual evaluation and automated quantification

AI for Multistructure Incidental Findings and Mortality Prediction at Chest CT in Lung Cancer Screening.

Machine learning computational model to predict lung cancer using electronic medical records

Automated detection of lung nodules and coronary artery calcium using artificial intelligence on low-dose CT scans for lung cancer screening: accuracy and prognostic value

Predicting Benign, Preinvasive, and Invasive Lung Nodules on Computed Tomography Scans Using Machine Learning

Automated Classification of Lung Cancer Subtypes Using Deep Learning and CT-Scan Based Radiomic Analysis

Auto-machine learning for opportunistic thyroid nodule detection in lung cancer screening chest CT.

Prediction of adenocarcinoma among other subtypes of lung cancer from CT using deep learning.

A machine learning-based PET/CT model for automatic diagnosis of early-stage lung cancer

Artificial-intelligence-based computed tomography histogram analysis predicting tumor invasiveness of lung adenocarcinomas manifesting as radiological part-solid nodules

Transparent Machine Learning Models to Diagnose Suspicious Thoracic Lesions Leveraging CT Guided Biopsy Data

Clinical, Conventional CT and Radiomic Feature-Based Machine Learning Models for Predicting ALK Rearrangement Status in Lung Adenocarcinoma Patients

Machine Learning Automatically Detects COVID-19 using Chest CTs in a Large Multicenter Cohort

Automatically Predicting Lung Adenocarcinoma Invasiveness

Prediction of pathologic stage in non-small cell lung cancer using machine learning algorithm based on CT image feature analysis

An Innovative Method for Lung Cancer Identification Using Machine Learning Algorithms

Application value of the automated machine learning model based on modified CT index combined with serological indices in the early prediction of lung cancer

CT Quantification of Interstitial Lung Abnormality and Interstitial Lung Disease: From Technical Challenges to Future Directions