Lung cancer is the leading cause of cancer-related deaths in Taiwan, with a
round 10,000 new cases annually. Sixty percent of diagnoses are in the ad
vanced stage, necessitating timely and precise treatment decisions. While
targeted therapies and immunotherapy have made progress, the five-year
survival rate for lung cancer patients remains low at around 15%. The proc
ess of selecting first-line treatment options based on pathology reports is
complex and time-consuming. To address these challenges, Taipei Medica
l University's interdisciplinary team has developed an innovative approa
ch using high-quality clinical data, natural language processing, and gene
rative AI. Their system can extract key features from pathology reports, pr
ovide treatment recommendations within seconds, and match patients tosuitable clinical trials. The team's BERT and GPT models analyze complex
reports and generate pathology summaries while ensuring data security.
By utilizing multimodal prediction methods, they can generate personaliz
ed treatment recommendations based on individual patient features. The
system incorporates the experiences of multiple doctors to produce highl
y accurate predictions, aiding decision-making between doctors and patie
nts. For patients with limited options, the model can match them with nea
rby new drug trial institutions, potentially extending their survival. The tec
hnology has applied for patents and offers a streamlined process, enhanci
ng medical decision-sharing and providing personalized and precise treat
ment options. It also improves healthcare quality in remote areas and add
resses manpower shortages. The application of AI in precision medicine el
evates Taiwan's healthcare standards.