This new method of assessment is more effective in predicting chemotherapy response as compared to previous studies.
According to the World Cancer Research Fund International, gastric cancer is the 5th most common cancer in the world. Early stages of the cancer can be easily overlooked as they are subtle and non-specific, much like those of long-term conditions like gastritis and stomach ulcers. Due to this, 80 to 90 per cent of individuals with stomach cancer are already in an advanced stage when they receive their first diagnosis. Currently, surgery is still the major treatment for advanced gastric cancer despite its poor prognosis and low five-year survival rate.
Neoadjuvant chemotherapy, which has grown in popularity among patients and surgeons in recent years, has helped patients with advanced gastric cancer have a better prognosis. However, roughly 30 per cent of advanced gastric cancer patients are unable to benefit from neoadjuvant chemotherapy and must deal with the danger of disease progression, further physical damage, and expensive treatment.
Although postoperative histopathology is the gold standard for assessing the effectiveness of neoadjuvant chemotherapy, it is unable to help in the optimisation of cancer therapy regimens. Therefore, it is essential to correctly identify advanced gastric cancer patients who have neoadjuvant chemotherapy resistance before treatment.
Recently, a team of researchers from the Suzhou Institute of Biomedical Engineering and Technology (SIBET) of the Chinese Academy of Sciences collaborated with the Shanxi Provincial Cancer Hospital to develop an artificial intelligence-based method for predicting the effectiveness of neoadjuvant chemotherapy for advanced gastric cancer. Based on intelligent technology for computing medical images, the team suggested a new efficacy prediction method that can address the issue of detecting advanced gastric cancer patients with neoadjuvant chemotherapy resistance.
To build the end-to-end neoadjuvant chemotherapy efficacy prediction model, the team used the ResNet-50 neural network architecture to automatically extract high-dimensional features from tumour images, combining spatial features of the tumour by using a multi-channel image input strategy and the boundary information of tumour to guide the network focus on the lesion area.
They also used computed tomography (CT) scans of 633 patients from three hospitals for model training and validation.
Compared to previous studies, the results obtained demonstrated that the suggested model is the most effective end-to-end neoadjuvant chemotherapy response prediction model, with high prediction accuracy and strong generalisation.
Moreover, to further demonstrate the interpretability of the model, the researchers used a visual technique to quantify the correspondence between tumour images and treatment resistance. The activation region of the tumour model in the CT image is not consistent, providing a reference for recognising the implicit link between tumour heterogeneity and treatment resistance. [APBN]
Source: Zhang et al. (2022). Deep learning predicts resistance to neoadjuvant chemotherapy for locally advanced gastric cancer: a multicenter study. Gastric Cancer, 1-10.