Early Detection, Prognosis Prediction, and Recurrence Monitoring of Lung Cancer Possible with a Single Urine Sample: Highly Accurate Lung Cancer Detection Using Urinary MicroRNA and AI
Craif Inc., a bio-AI startup, in collaboration with Tokyo Jikei University School of Medicine and Higashiosaka City Medical Center, has developed a non-invasive testing platform that enables early detection, prognosis prediction, and recurrence monitoring of lung cancer using urinary microRNA and AI. This research, published in "npj Precision Oncology" on April 18, 2026, allows for home urine collection and is expected to be applied to widespread screening.
📋 Article Processing Timeline
- 📰 Published: April 30, 2026 at 19:00
- 🔍 Collected: April 30, 2026 at 10:31
- 🤖 AI Analyzed: April 30, 2026 at 21:11 (10h 40m after Collected)
Bio-AI startup Craif Inc. (Location: Shinjuku-ku, Tokyo; Representative Director CEO: Ryuichi Onose; hereinafter "Craif"), in a joint research project with Associate Professor Yu Fujita, Head of the Department of Next-Generation Drug Discovery Research, Comprehensive Medical Science Research Center, The Jikei University School of Medicine, and Dr. Takashi Nojiri, Head of the Department of Thoracic Surgery, Higashiosaka City Medical Center, has demonstrated the potential for highly accurate early detection of lung cancer through a non-invasive urine test that requires no blood collection or hospital visits. This is achieved by analyzing microRNAs contained in urine using AI (artificial intelligence). The research also showed the possibility of predicting postoperative recurrence risk from preoperative urine and monitoring recurrence through changes in urinary microRNA levels after surgery. These research findings were published in the academic journal "npj Precision Oncology" on April 18, 2026.
(Figure 1) Overview of this research
■ Highlights of the Announcement
Lung cancer, difficult to detect early: Lung cancer is the leading cause of cancer deaths worldwide. A major challenge is that early symptoms are often subtle, and many patients are diagnosed at an advanced stage.
Multi-institutional collaborative research conducted: Urine samples from 278 lung cancer patients (approximately half of whom were in early stages) and 213 non-cancer subjects were analyzed.
Highly accurate detection of early lung cancer with urinary microRNA test: By enriching and analyzing microRNAs derived from extracellular vesicles in urine and constructing a machine learning model, early lung cancer was detected with high accuracy (test set: AUC 0.941, early stage sensitivity 88.2%, specificity 87.0%).
Potential for recurrence monitoring through changes in urinary microRNA before and after surgery: 12 types of microRNAs were identified that are highly expressed in lung cancer patients, decrease after surgery, and tend to re-elevate upon recurrence.
Stratification of postoperative recurrence risk with a 3-microRNA panel: A panel of three microRNAs (hsa-miR-181a-5p, hsa-miR-185-5p, hsa-miR-934) showed a significant difference in recurrence-free survival between high-risk and low-risk groups (HR=8.3).
Single urinary microRNA assay platform addresses multiple phases of lung cancer care: This platform demonstrated the potential to comprehensively cover different phases of lung cancer management—early detection, prognosis prediction, and recurrence monitoring—with a single testing infrastructure.
Non-invasive test allowing home urine collection: Since it uses urine, no blood collection is required, and samples can be collected at home. This is expected to be utilized for widespread population screening and in areas with limited access to medical facilities.
■ Overview of this research
Lung cancer is the most common cancer in men and the second most common in women, and it is the leading cause of cancer deaths worldwide. Early symptoms are often almost nonexistent, and by the time it is discovered, it is often already at a stage where surgical removal is difficult, making early detection key to treatment. Furthermore, even for patients who undergo surgery at an early stage, there remains a risk of postoperative recurrence. Current blood tumor markers (such as CEA and CYFRA21-1) have low sensitivity for early lung cancer, and there is a demand for the development of more accurate non-invasive testing methods.
Therefore, the research team focused on extracellular vesicles and microRNAs, which are deeply involved in the activity of cancer cells. Cells transmit information to distant cells in the body via microRNAs contained in extracellular vesicles, and cancer cells are also thought to actively utilize this mechanism. By enriching extracellular vesicles in urine, the team focused on efficiently extracting and analyzing large amounts of microRNAs from urine and worked on developing a non-invasive lung cancer test.
In this study, urine samples were collected from 278 lung cancer patients (approximately half of whom were in early stages [Stage 0/I]) and 213 non-cancer subjects from four institutions (Jikei University Hospital, Higashiosaka City Medical Center, Hokuto Hospital, Omiya City Clinic) and analyzed using Small-RNA sequencing and machine learning. The constructed lung cancer detection model showed high performance with an AUC of 0.942 in the training set and an AUC of 0.941 in the independent test set. Specifically, for early-stage lung cancer, a sensitivity of 88.2% and a specificity of 87.0% were achieved in the test set (Figure 2). It was also confirmed that the prediction score was not affected by background factors such as age, sex, BMI, and smoking history.
(Figure 2) ROC curve of the lung cancer detection model (left) and distribution of stage-specific prediction scores in the test set (right)
As an application to recurrence monitoring, comparative analysis of urine samples before and after surgery identified 12 types of microRNAs that are highly expressed in the lung cancer group and decrease after surgery. These microRNAs showed a tendency to re-elevate upon recurrence, suggesting their potential as biomarkers reflecting the presence of tumors.
Furthermore, for the purpose of predicting postoperative recurrence risk, a Cox regression analysis was performed on 76 lung cancer patients in Stage I/II.
(Figure 1) Overview of this research
■ Highlights of the Announcement
Lung cancer, difficult to detect early: Lung cancer is the leading cause of cancer deaths worldwide. A major challenge is that early symptoms are often subtle, and many patients are diagnosed at an advanced stage.
Multi-institutional collaborative research conducted: Urine samples from 278 lung cancer patients (approximately half of whom were in early stages) and 213 non-cancer subjects were analyzed.
Highly accurate detection of early lung cancer with urinary microRNA test: By enriching and analyzing microRNAs derived from extracellular vesicles in urine and constructing a machine learning model, early lung cancer was detected with high accuracy (test set: AUC 0.941, early stage sensitivity 88.2%, specificity 87.0%).
Potential for recurrence monitoring through changes in urinary microRNA before and after surgery: 12 types of microRNAs were identified that are highly expressed in lung cancer patients, decrease after surgery, and tend to re-elevate upon recurrence.
Stratification of postoperative recurrence risk with a 3-microRNA panel: A panel of three microRNAs (hsa-miR-181a-5p, hsa-miR-185-5p, hsa-miR-934) showed a significant difference in recurrence-free survival between high-risk and low-risk groups (HR=8.3).
Single urinary microRNA assay platform addresses multiple phases of lung cancer care: This platform demonstrated the potential to comprehensively cover different phases of lung cancer management—early detection, prognosis prediction, and recurrence monitoring—with a single testing infrastructure.
Non-invasive test allowing home urine collection: Since it uses urine, no blood collection is required, and samples can be collected at home. This is expected to be utilized for widespread population screening and in areas with limited access to medical facilities.
■ Overview of this research
Lung cancer is the most common cancer in men and the second most common in women, and it is the leading cause of cancer deaths worldwide. Early symptoms are often almost nonexistent, and by the time it is discovered, it is often already at a stage where surgical removal is difficult, making early detection key to treatment. Furthermore, even for patients who undergo surgery at an early stage, there remains a risk of postoperative recurrence. Current blood tumor markers (such as CEA and CYFRA21-1) have low sensitivity for early lung cancer, and there is a demand for the development of more accurate non-invasive testing methods.
Therefore, the research team focused on extracellular vesicles and microRNAs, which are deeply involved in the activity of cancer cells. Cells transmit information to distant cells in the body via microRNAs contained in extracellular vesicles, and cancer cells are also thought to actively utilize this mechanism. By enriching extracellular vesicles in urine, the team focused on efficiently extracting and analyzing large amounts of microRNAs from urine and worked on developing a non-invasive lung cancer test.
In this study, urine samples were collected from 278 lung cancer patients (approximately half of whom were in early stages [Stage 0/I]) and 213 non-cancer subjects from four institutions (Jikei University Hospital, Higashiosaka City Medical Center, Hokuto Hospital, Omiya City Clinic) and analyzed using Small-RNA sequencing and machine learning. The constructed lung cancer detection model showed high performance with an AUC of 0.942 in the training set and an AUC of 0.941 in the independent test set. Specifically, for early-stage lung cancer, a sensitivity of 88.2% and a specificity of 87.0% were achieved in the test set (Figure 2). It was also confirmed that the prediction score was not affected by background factors such as age, sex, BMI, and smoking history.
(Figure 2) ROC curve of the lung cancer detection model (left) and distribution of stage-specific prediction scores in the test set (right)
As an application to recurrence monitoring, comparative analysis of urine samples before and after surgery identified 12 types of microRNAs that are highly expressed in the lung cancer group and decrease after surgery. These microRNAs showed a tendency to re-elevate upon recurrence, suggesting their potential as biomarkers reflecting the presence of tumors.
Furthermore, for the purpose of predicting postoperative recurrence risk, a Cox regression analysis was performed on 76 lung cancer patients in Stage I/II.