# Deep Learning Reconstruction Enhances Lung Cancer CT Imaging

## TL;DR

Deep learning improves CT image quality for lung cancer, helping doctors make better treatment decisions.

## Contribution

DLR combined with UHRCT improves imaging of apical lung tumors, enabling better treatment planning.

## Key findings

- DLR reduces noise and streak artifacts in UHRCT images of lung apex tumors.
- DLR enables clearer visualization of tumor-chest wall relationships.
- DLR aids in determining appropriate treatment for anatomically complex lung tumors.

## Abstract

A 70-year-old man was referred for the surgical treatment of a right upper lobe lung adenocarcinoma. Preoperative CT revealed a tumor approximately 26 mm in size; however, the relationship between the tumor and the adjacent chest wall could not be assessed owing to noise and streak artifacts, typical of the lung apex. Ultra-high-resolution CT (UHRCT) was performed using an Aquilion Precision scanner (Canon Medical Systems; 1,792 channels per row, 0.25 mm × 160 rows, 1,024 matrix) to improve diagnostic accuracy. Images were reconstructed at 0.25-mm slice thickness using a deep learning-based reconstruction (DLR). Compared with conventional filtered back projection, the DLR images demonstrated markedly reduced noise and streak artifacts from the shoulder and clavicle, substantially improving image quality. On mediastinal window settings, tumor invasion into the superior chest wall was visualized. We thus inferred surgical resection as inappropriate; therefore, systemic chemotherapy was selected. This case demonstrates that UHRCT combined with DLR is useful for evaluating apical lung tumors that are difficult to assess using conventional CT. High-quality images provided clearer delineation of the relationship between the tumor and adjacent structures, contributing to treatment planning. DLR is a promising diagnostic approach for anatomically challenging regions such as the lung apex.