Background
Image recognition using artificial intelligence with deep learning through convolutional neural networks (CNNs) has dramatically improved and been increasingly applied to medical fields ...for diagnostic imaging. We developed a CNN that can automatically detect gastric cancer in endoscopic images.
Methods
A CNN-based diagnostic system was constructed based on Single Shot MultiBox Detector architecture and trained using 13,584 endoscopic images of gastric cancer. To evaluate the diagnostic accuracy, an independent test set of 2296 stomach images collected from 69 consecutive patients with 77 gastric cancer lesions was applied to the constructed CNN.
Results
The CNN required 47 s to analyze 2296 test images. The CNN correctly diagnosed 71 of 77 gastric cancer lesions with an overall sensitivity of 92.2%, and 161 non-cancerous lesions were detected as gastric cancer, resulting in a positive predictive value of 30.6%. Seventy of the 71 lesions (98.6%) with a diameter of 6 mm or more as well as all invasive cancers were correctly detected. All missed lesions were superficially depressed and differentiated-type intramucosal cancers that were difficult to distinguish from gastritis even for experienced endoscopists. Nearly half of the false-positive lesions were gastritis with changes in color tone or an irregular mucosal surface.
Conclusion
The constructed CNN system for detecting gastric cancer could process numerous stored endoscopic images in a very short time with a clinically relevant diagnostic ability. It may be well applicable to daily clinical practice to reduce the burden of endoscopists.
The prognosis of esophageal cancer is relatively poor. Patients are usually diagnosed at an advanced stage when it is often too late for effective treatment. Recently, artificial intelligence (AI) ...using deep learning has made remarkable progress in medicine. However, there are no reports on its application for diagnosing esophageal cancer. Here, we demonstrate the diagnostic ability of AI to detect esophageal cancer including squamous cell carcinoma and adenocarcinoma.
We retrospectively collected 8428 training images of esophageal cancer from 384 patients at the Cancer Institute Hospital, Japan. Using these, we developed deep learning through convolutional neural networks (CNNs). We also prepared 1118 test images for 47 patients with 49 esophageal cancers and 50 patients without esophageal cancer to evaluate the diagnostic accuracy.
The CNN took 27 seconds to analyze 1118 test images and correctly detected esophageal cancer cases with a sensitivity of 98%. CNN could detect all 7 small cancer lesions less than 10 mm in size. Although the positive predictive value for each image was 40%, misdiagnosing shadows and normal structures led to a negative predictive value of 95%. The CNN could distinguish superficial esophageal cancer from advanced cancer with an accuracy of 98%.
The constructed CNN system for detecting esophageal cancer can analyze stored endoscopic images in a short time with high sensitivity. However, more training would lead to higher diagnostic accuracy. This system can facilitate early detection in practice, leading to a better prognosis in the near future.
Display omitted
Background
Determining the depth of invasion of early stage colorectal cancer has been emphasized as a means of improving endoscopic diagnostic accuracy. Recent studies have focused on other ...pathological risk factors for lymph node metastasis (LNM). We investigated the significance of depth of invasion and predictive properties of other endoscopic findings.
Methods
We retrospectively investigated 846 patients with submucosal invasive (T1) colorectal cancer who received an accurate pathological diagnosis and were treated between January 2005 and December 2016. Pathological risk factors associated with LNM were reviewed. We divided patients into groups: low-risk T1 colorectal cancer (LRC; no risk factors) and high-risk T1 colorectal cancer (HRC; exhibiting lymphovascular invasion, tumor budding grade of 2/3, and/or poor differentiation) and studied predictive endoscopic factors for HRC.
Results
Significant risk factors for LNM in multivariate analysis were lymphovascular invasion odds ratio (OR) 8.09; 95% confidence interval (CI) 3.84–17.1, tumor budding (OR 1.89; 95% CI 1.09–3.29), and histological differentiation (OR 2.09; 95% CI 1.12–3.89). The LNM-positive rate with only deep submucosal invasion was 1.6%. Significant predictive factors for HRC in multivariate analysis identified rectal tumor location (OR 1.92; 95% CI 1.35 –2.72, depression (OR 2.73; 95% CI 1.96 –3.80), protuberance within the depression (OR 2.58; 95% CI 1.39– 4.78), expansiveness (OR 2.39; 95% CI 1.27– 4.50), and loss of mucosal patterns (OR 1.90; 95% CI 1.20 –3.01) as significant factors.
Conclusions
Rectal tumor location, depression, protuberance within the depression, expansiveness, and loss of mucosal patterns could be predictive factors for HRC.
Background
Early detection of early gastric cancer (EGC) allows for less invasive cancer treatment. However, differentiating EGC from gastritis remains challenging. Although magnifying endoscopy with ...narrow band imaging (ME-NBI) is useful for differentiating EGC from gastritis, this skill takes substantial effort. Since the development of the ability to convolve the image while maintaining the characteristics of the input image (convolution neural network: CNN), allowing the classification of the input image (CNN system), the image recognition ability of CNN has dramatically improved.
Aims
To explore the diagnostic ability of the CNN system with ME-NBI for differentiating between EGC and gastritis.
Methods
A 22-layer CNN system was pre-trained using 1492 EGC and 1078 gastritis images from ME-NBI. A separate test data set (151 EGC and 107 gastritis images based on ME-NBI) was used to evaluate the diagnostic ability accuracy, sensitivity, positive predictive value (PPV), and negative predictive value (NPV) of the CNN system.
Results
The accuracy of the CNN system with ME-NBI images was 85.3%, with 220 of the 258 images being correctly diagnosed. The method’s sensitivity, specificity, PPV, and NPV were 95.4%, 71.0%, 82.3%, and 91.7%, respectively. Seven of the 151 EGC images were recognized as gastritis, whereas 31 of the 107 gastritis images were recognized as EGC. The overall test speed was 51.83 images/s (0.02 s/image).
Conclusions
The CNN system with ME-NBI can differentiate between EGC and gastritis in a short time with high sensitivity and NPV. Thus, the CNN system may complement current clinical practice of diagnosis with ME-NBI.
Since the indications for endoscopic submucosal dissection have been expanded to include undifferentiated-type early gastric cancers, improvements in preoperative diagnostic ability have been an area ...of research. There are also concerns about the impact on the diagnosis of
infection. Based on our previous studies, in undifferentiated-type early gastric cancers, magnifying endoscopy with narrow-band imaging is useful for delineating the demarcation regardless of the tumor size. Additionally, inflammatory cell infiltration appears to be a cause of misdiagnosis, suggesting that the resolution of inflammation could contribute to the accurate diagnosis of demarcations. As such, the accuracy of demarcation in eradicated and uninfected cases is higher than that in non-eradicated cases. The common features of the endoscopic findings were discoloration under white-light imaging and a predominance of sites in the lower and middle regions. The uninfected group was characterized by smaller tumor size, flat type, more extended intervening parts in magnifying endoscopy with narrow-band imaging, and pure signet ring cell carcinoma. In contrast, the eradication and non-eradication groups were characterized by larger tumor size, depressed type, and wavy microvessels in magnifying endoscopy with narrow-band imaging. In this comprehensive review, as described above, we discuss the diagnosis of demarcation of undifferentiated-type early gastric cancers, undifferentiated-type early gastric cancers that developed following
eradication, and
-uninfected undifferentiated-type early gastric cancers, with a focus on studies with self-examination and endoscopic findings and describe the future direction.
Objectives
Detecting early gastric cancer is difficult, and it may even be overlooked by experienced endoscopists. Recently, artificial intelligence based on deep learning through convolutional ...neural networks (CNNs) has enabled significant advancements in the field of gastroenterology. However, it remains unclear whether a CNN can outperform endoscopists. In this study, we evaluated whether the performance of a CNN in detecting early gastric cancer is better than that of endoscopists.
Methods
The CNN was constructed using 13,584 endoscopic images from 2639 lesions of gastric cancer. Subsequently, its diagnostic ability was compared to that of 67 endoscopists using an independent test dataset (2940 images from 140 cases).
Results
The average diagnostic time for analyzing 2940 test endoscopic images by the CNN and endoscopists were 45.5 ± 1.8 s and 173.0 ± 66.0 min, respectively. The sensitivity, specificity, and positive and negative predictive values for the CNN were 58.4%, 87.3%, 26.0%, and 96.5%, respectively. These values for the 67 endoscopists were 31.9%, 97.2%, 46.2%, and 94.9%, respectively. The CNN had a significantly higher sensitivity than the endoscopists (by 26.5%; 95% confidence interval, 14.9–32.5%).
Conclusion
The CNN detected more early gastric cancer cases in a shorter time than the endoscopists. The CNN needs further training to achieve higher diagnostic accuracy. However, a diagnostic support tool for gastric cancer using a CNN will be realized in the near future.
Gastric cancer in young adults has been pointed out to comprise a subgroup associated with distinctive clinicopathological features, including an equal gender distribution, advanced disease, and ...diffuse‐type histology. Comprehensive molecular analyses of gastric cancers have led to molecular‐based classifications and to specific and effective treatment options. The molecular traits of gastric cancers in young adults await investigations, which should provide a clue to explore therapeutic strategies. Here, we studied 146 gastric cancer patients diagnosed at the age of 40 years or younger at the Cancer Institute Hospital (Tokyo, Japan). Tumor specimens were examined for Helicobacter pylori infection, Epstein‐Barr virus positivity, and for the expression of mismatch repair genes to indicate microsatellite instability. Overexpression, gene amplifications, and rearrangements of 18 candidate driver genes were examined by immunohistochemistry and FISH. Although only a small number of cases were positive for Epstein‐Barr virus and microsatellite instability (n = 2 each), we repeatedly found tumors with gene fusion between a tight‐junction protein claudin, CLDN18, and a regulator of small G proteins, ARHGAP, in as many as 22 cases (15.1%), and RNA sequencing identified 2 novel types of the fusion. Notably, patients with the CLDN18‐ARHGAP fusion revealed associations between aggressive disease and poor prognosis, even when grouped by their clinical stage. These observations indicate that a fusion gene between CLDN18 and ARHGAP is enriched in younger age‐onset gastric cancers, and its presence could contribute to their aggressive characteristics.
In our gastric cancer in young adult cohort, we found enrichment of fusion genes between CLDN18 and ARHGAP. The positivity of the CLDN18‐ARHGAP fusion relates to advanced disease and poor prognosis, indicating its clinical relevance.
Background
Management strategies for primary non-ampullary duodenal adenocarcinoma (NADAC) in early stage are not well established given its low incidence. This study aimed to elucidate ...clinicopathological features of early NADAC, including risk for lymph nodal metastasis (LNM).
Methods
In total, 166 patients with early NADAC underwent initial treatment at our institution between 2006 and 2019, of whom 153 had intramucosal (M-) and 13 had submucosal (SM-) NADAC. These endoscopic and pathological features were retrospectively analyzed. Risk factors for LNM were evaluated in 46 early NADAC patients who underwent surgery with lymph node dissection.
Results
Compared with M-NADAC, SM-NADAC was significantly more frequently located at the proximal side of the papilla, with mixed elevated and depressed macroscopic type, histologically poorly differentiated tumor and lymphovascular invasion (LVI) (85% vs. 47%,
P
= 0.009; 54% vs. 5%,
P
< 0.001; 23% vs. 0%,
P
< 0.001; and 46% vs. 0%,
P
< 0.001, respectively). The frequency of LNM was significantly higher in SM-NADAC than in M-NADAC (5/12, 42% vs. 0/34, 0%;
P
< 0.001). In SM-NADAC, the frequency of LNM was higher in poorly differentiated than in well to moderately differentiated tumors (3/3, 100% vs. 2/9, 22%) and higher in tumors with LVI than in those without LVI (3/5, 60% vs. 2/7, 29%). Regarding invasion depth, 2 of 4 patients with SM invasion (400 ≤ × < 500 µm) showed LNM. However, in this study, no patients developed very shallow SM invasion (0 < × < 400 µm).
Conclusions
SM-NADAC showed high LNM risk. Surgical treatment with regional lymph node dissection is recommended as a treatment strategy for SM-NADAC.
Background
Colorectal endoscopic submucosal dissection (ESD) requires advanced endoscopic skill. For safer and more reliable ESD implementation, various traction devices have been developed in recent ...years. The purpose of this research was to evaluate whether an ESD training program using a traction device (TD) would contribute to the improvement of trainees’ skill acquisition.
Methods
The differences in treatment outcomes and learning curves by the training program were compared before and after the introduction of TD (control group: January 2014 to March 2016; TD group: April 2016 to June 2018).
Results
A total of 316 patients were included in the analysis (TD group: 202 cases; control group: 114 cases). The number of cases required to achieve proficiency in ESD techniques was 10 in the TD group and 21 in the control group. Compared to the control group, the TD group had a significant advantage in ESD self-completion rate (73.8% vs. 58.8%), dissection speed (19.5 mm
2
/min vs. 15.9 mm
2
/min), en bloc resection rate (100% vs. 90%), and R0 resection rate (96% vs. 83%).
Conclusions
The rate of colorectal ESD self-completion by trainees improved immediately after the start of the training program using a traction device compared to the conventional method, and the dissection speed tended to increase linearly with ESD experience. We believe that ESD training using a traction device will help ESD techniques to be performed safely and reliably among trainees.
Background
While endoscopic submucosal dissection (ESD) is recognized as a minimally invasive standard treatment for differentiated early gastric cancers (EGCs), it has not been indicated for ...undifferentiated EGC (UD-EGC) because of a relatively high risk of lymph node metastasis (LNM). However, patients with surgically resected mucosal (cT1a) UD-EGC ≤ 2 cm in size with no lymphovascular invasion or ulceration are reported to be at a very low risk of LNM. This multicenter, single-arm, confirmatory trial was conducted to evaluate the efficacy and safety of ESD for UD-EGC.
Methods
The key eligibility criteria were endoscopically diagnosed cT1a/N0/M0, single primary lesion, size ≤ 2 cm, no ulceration and histologically proven components of undifferentiated adenocarcinoma on biopsy. Based on the histological findings after ESD, additional gastrectomy was indicated if the criteria for curative resection were not satisfied. The subjects of the primary analysis were patients with UD-EGC as the dominant component. The primary endpoint was 5-year overall survival (OS) of patients with UD-EGC.
Results
Three hundred 46 patients were enrolled from 49 institutions. The proportion of
en bloc
resection was 99%. No ESD-related Grade 4 adverse events were noted. Delayed bleeding and intraoperative and delayed perforation occurred in 25 (7.3%), 13 (3.8%), and 6 (1.7%) patients, respectively. Among the 275 patients who were the subjects of the primary analysis, curative resection was achieved in 195 patients (71%), and 5-year OS was 99.3% (95% CI: 97.1–99.8).
Conclusions
ESD can be a curative and less invasive treatment for UD-EGC for patients meeting the eligibility criteria of this study.