Five Pivotal AVIEW Publications to Follow in 2024

Hello. I'm Donghoon Yu from the AI1T Lab.
I've been introducing papers that used our company's products through Slack from time to time, and this time I'd like to gather and organize some of the notable papers published in 2024.

The number of papers mentioning aview, our product, is increasing every year. This year, a total of 108 papers were published in 2024, which is a significant increase from 80 in 2023. These statistics show that aview is widely used around the world.

Below are five major examples of papers published in 2024.

Feasibility of AI as First Reader in the 4-IN-THE-LUNG-RUN Lung Cancer Screening Trial: Impact on Negative-Misclassifications and Clinical Referral Rate

---

This paper, published by iDNA, evaluated the possibility of using AI software (AVIEW-LCS) as the first reader in the 4-IN-THE-LUNG-RUN experiment. This study compared the performance of AI and radiologists to analyze the impact on the identification of negative cases and clinical referral rates. Of the 3,678 participants, AI performed better than radiologists, with 31 cases (0.8%) of negative misclassification (NM) recorded, compared to 407 cases (11.1%). In addition, the NM Referral rate of AI was 2.9%, which was significantly lower than that of radiologists (11.8%). AI showed the potential to reduce workload by 71.2% while maintaining diagnostic safety. This study highlights the potential of AI to improve diagnostic workflows and reduce the burden on radiologists. Here, NM refers to a false negative classification in the first reading, and NM Referrel refers to a false negative classification in the final reading. It also mentions false positives (PM), which are 5.7% for AI, which is higher than the 0.5% for radiologists, but the load caused by this is said to be minor and further research is needed.

 

Coronary calcium score and emphysema extent on different CT radiation dose protocols in lung cancer screening

---

The paper, published by the University of Parma, investigated the possibility of automatically quantifying coronary artery calcification (CAC) and emphysema using low-dose (LDCT) and ultra-low-dose (ULDCT) CT scans in lung cancer screening. 361 participants were scanned using AVIEW software, and CAC measurements showed a high degree of agreement (ICC = 0.86) and excellent overlap (84%) between LDCT and ULDCT. On the other hand, **emphysema measurements showed a moderate degree of agreement (ICC = 0.57)**, and an overestimation trend was observed in ULDCT. The results of the study show that ULDCT is suitable for quantifying CAC, suggesting the possibility of expanding low-dose lung cancer screening to cardiovascular prevention. This study shows that AVIEW CAC can also be applied to ULDCT.


Comparison of AI software tools for automated detection, quantification and categorization of pulmonary nodules in the HANSE LCS trial

---

This is a paper that compares Nodule CAD with Siemens products and was researched through the HANSE pilot project. Professor Claussen is included as an author. This study compared AI-based software **Aview (Coreline Soft, South Korea, S1)** and ChestCTExplore (Siemens Healthineers, Germany, S2) in the HANSE Lung Cancer Screening (LCS) study. The study analyzed 946 low-dose CT scans and compared the AI results with the readings of radiologists. The volume correlation (r > 0.95) between the two AI tools was high, but S2 tended to measure a larger volume than S1. S1 performed better than S2 in sensitivity, positive predictive value (PPV), and concordance with the final reading of the radiologist. In addition, the performance difference was clear, with S1 showing a 75% match rate in the Lung-RADS classification, while S2 recorded a 55% match rate. In particular, the Lung-RADS score varied depending on the AI tool in 38% of participants, which may affect patient treatment decisions. The study emphasized the need for consistency between AI tools and supervision by radiologists.


Artificial intelligence system for identification of overlooked lung metastasis in abdominopelvic computed tomography scans of patients with malignancy

---

This is a paper that confirms that the application of Lung Nodule CAD to abdominal CT scans can help to find missed nodules. The authors include two people, Jae-Yeon Lee and Boo-Lim Choi. This study demonstrated that the AI system AVIEW can successfully identify lung metastases missed by abdominal CT scans. The AI showed an AUC of 0.911 and a sensitivity of up to 92.3%. The FP was 27.6%, but it could be significantly reduced to 2.4-12.6% through a review of radiologists. Using AI as a second reader showed that it could reduce diagnostic errors and minimize unnecessary reviews.

Development and multi-institutional validation of estimating forced vital capacity in pulmonary fibrosis using quantitative chest CT data

---

This is not a journal article, but an academic paper presented at RSNA 2024, written by Jang Ryung-woo, the lead author. It is about how to estimate FVC (forced vital capacity) using AVIEW lung texture analysis results. It showed a correlation coefficient of 0.68 in the entire data, including IPF patients and non-IPF patients, and a correlation coefficient of 0.77 in the non-IPF patient group, showing the possibility of being used to evaluate the progression of the disease based on imaging alone.

 

That concludes our introduction of five particularly noteworthy papers published in 2024. It is impressive that AVIEW is being used in various studies this year and its reliability and performance are being recognized. In particular, it is proving its clinical value by showing significant results in various fields, including AI-based lung cancer screening, analysis of cardiovascular and respiratory diseases, and abdominal CT.

I hope that we will all continue to work together to ensure that our company's technology will continue to provide practical value in the medical field through continued research and cooperation.