Another milestone for Senticlab in lung cancer recognition

The Senticlab team has developed a new solution based on artificial intelligence to automatically recognize tumoral nodules from chest radiographs, obtaining one of the best results in the NODE21 competition.

Medical Imaging Analysis and Diagnostics Machine Learning Artificial Intelligence

Lung cancer is a serious illness causing the greatest number of cancer deaths worldwide. Symptoms of lung cancer typically occur at an advanced stage of the disease, when treatment has a reduced chance of success. For these reasons, early diagnosis is crucial in order to reduce mortality rates from lung cancer.

To this end, screening processes are based on the identification of pulmonary nodules, which can be detected by analyzing chest radiograph and are visible well before clinical symptoms or signs emerge. Since exams requiring chest radiographs are quite common, pulmonary nodules are frequently encountered as incidental findings in patients undergoing routine examination or CXR imaging for issues unrelated to lung cancer.

This context has always been of particular interest for Senticlab, a society owned by synbrAIn, by means of which we have developed several AI solution in the field of healthcare and diagnosis support. Since the identification of pulomnar nodules was already part of the synbrAIn product development roadmap, the data science team of Senticlab joined the NODE21 competition, which goal is to compare various solution to identify tumoral nodules from radiographic chest images.

Polmonar nodes detection

Senticlab has developed a solution based on YOLO, a modern computer vision algorithm which is especially effective for object detection tasks. By applying a properly modified version of YOLO on chest radiographs, it is possible to identify tumoral nodules with an accuracy of 82.75%. This result, based on a dataset of 4882 frontal chest radiographs, allowed Senticlab to be ranked second among all the other solutions participating in NODE21 for the detection task, just over a percentage point from the winning solution (MTEC, Hamburg University of Technology), and distancing the solution proposed by the prestigious UCLA by two percentage points.

Beyond the results, this is the umpteenth proof of the effectiveness of AI in supporting medical diagnosis. Together with Senticlab, synbrAIn has developed several applications that have already been placed on the market within the MS HUMANAID platform for the automatic identification of pneumonia from medical images. In addition to the identification of pulmonary nodules, sybrAIn and Senticlab are working to extend the functionality of MS HUMANAID / AIRX with CT analysis and the identification of Pneumothorax.

We are convinced that the use of these technologies will continue to increase more and more, finding practical applications especially in the context of diagnosis support, implementing that human-machine cooperation on which our entire company philosophy is based.