Application of Artificial Intelligence in Lung Cancer Diagnosis Based on Chest CT Examination

A lung cancer diagnosis is of paramount importance in modern medicine, as this disease is among the most common causes of death in many countries around the world. Early detection of lung cancer fundamentally determines the survival chances of patients, as the likelihood of recovery is much higher in the first stage than in later phases. However, the disease often only shows symptoms in the later stages, making early diagnosis challenging. Traditional screening methods, such as chest X-rays, are not always effective enough to detect early-stage abnormalities.

The Role of Artificial Intelligence in Lung Cancer Diagnostics

The development of artificial intelligence (AI) offers an opportunity to rethink diagnostic processes. The application of AI in radiological diagnostics not only speeds up workflows but also increases the accuracy of diagnoses. Through the latest research and developments, a Hungarian consortium, which includes Semmelweis University and Ulyssys Ltd., has developed an innovative lung cancer diagnostic solution. Its aim is to support lung cancer screening programs, thereby contributing to the early detection of the disease and improving recovery rates.

The Situation of Lung Cancer and Diagnostic Challenges

Lung cancer is one of the most common types of cancer and poses a serious problem for public health according to mortality statistics. Early-stage diagnosis of the disease is essential for improving patients’ quality of life and increasing survival rates. Traditional methods used for lung cancer screening, such as chest X-rays, are not always reliable, as early-stage abnormalities are often undetectable.

Recent studies indicate that low-dose CT scans (LDCT) offer significant advantages in lung cancer screening, especially among smokers. LDCT has been shown to reduce lung cancer mortality; however, the implementation of screening programs requires the evaluation of a large number of chest CT scans, which places a tremendous burden on radiologists. The European Union has made recommendations for the introduction of CT-based screening, but the available radiological resources are limited, necessitating urgent solutions to accelerate and enhance the efficiency of diagnostic processes.

The Role of Artificial Intelligence in Lung Cancer Diagnostics

The advancements in artificial intelligence, particularly in the field of deep learning (DL), provide significant advantages in radiological diagnostics. The partnership between Ulyssys Ltd. and Semmelweis University aims to develop a lung cancer diagnostic system that employs AI-based solutions to improve the effectiveness of lung cancer screening. AI is capable of automatically analyzing CT images, identifying suspicious lesions, and prioritizing cases awaiting evaluation.

Additionally, the system generates brief reports and summaries of the images, facilitating the work of radiologists. The 3D segmentations generated by AI allow for more accurate detection of suspicious lesions, thereby increasing the accuracy of diagnoses. The use of AI not only reduces diagnostic time but also saves significant resources for radiologists, enabling them to see more patients.

The developed AI system has demonstrated its effectiveness in practice, achieving a sensitivity exceeding 96% in an independent test cohort, while specificity also showed values above 80%. This performance indicates that the system can reliably identify the suspicion of lung cancer, thereby contributing to early diagnoses and saving patients’ lives.

Future Prospects and the Clinical Validation Process

During the development of the lung cancer diagnostic system, researchers paid attention not only to technological solutions but also to clinical applicability. Over 8000 low-dose chest CT scans were collected and analyzed within the project to enable AI to identify the signs of lung cancer as accurately as possible. Through research conducted by Semmelweis University, the system continuously improved by comparing tumor segmentations and manual evaluations.

As part of the developments, Ulyssys Ltd. created a software solution that allows for the uploading of CT images and the execution of diagnostic analyses. With this solution, radiologists can make their diagnostic decisions more easily. To facilitate market introduction, the company has initiated the process of obtaining ISO 13485 certification and CE/MDR certification to ensure the legal and administrative compliance of the system.

Through clinical validation studies, the developed system is also being tested in real clinical environments, which is essential for successful market introduction. The goal is for the AI-supported lung cancer diagnostic system to contribute to the early detection of lung cancer not only in Hungary but also internationally, thereby improving patients’ survival chances and quality of life.