Analyzing risk factors and constructing a predictive model for superficial esophageal carcinoma with submucosal infiltration exceeding 200 micrometers.

OBJECTIVE: Submucosal infiltration of less than 200 µm is considered an indication for endoscopic surgery in cases of superficial esophageal cancer and precancerous lesions. This study aims to identify the risk factors associated with submucosal infiltration exceeding 200 micrometers in early esophageal cancer and precancerous lesions, as well as to establish and validate an accompanying predictive model. METHODS: Risk factors were identified through least absolute shrinkage and selection operator (LASSO) and multivariate logistic regression. Various machine learning (ML) classification models were tested to develop and evaluate the most effective predictive model, with Shapley Additive Explanations (SHAP) employed for model visualization. RESULTS: Predictive factors for early esophageal invasion into the submucosa included endoscopic ultrasonography or magnifying endoscopy> SM1(P<0.001,OR = 3.972,95%CI 2.161-7.478), esophageal wall thickening(P<0.001,OR = 12.924,95%CI,5.299-33.96), intake of pickled foods(P=0.04,OR = 1.837,95%CI,1.03-3.307), platelet-lymphocyte ratio(P<0.001,OR = 0.284,95%CI,0.137-0.556), tumor size(P<0.027,OR = 2.369,95%CI,1.128-5.267), the percentage of circumferential mucosal defect(P<0.001,OR = 5.286,95%CI,2.671-10.723), and preoperative pathological type(P<0.001,OR = 4.079,95%CI,2.254-7.476). The logistic regression model constructed from the identified risk factors was found to be the optimal model, demonstrating high efficacy with an area under the curve (AUC) of 0.922 in the training set, 0.899 in the validation set, and 0.850 in the test set. CONCLUSION: A logistic regression model complemented by SHAP visualizations effectively identifies early esophageal cancer reaching 200 micrometers into the submucosa.

Interobserver agreement on line-field confocal optical coherence tomography image markers in keratinocyte carcinomas and precursor lesions.

Line-field confocal optical coherence tomography (LC-OCT) is a new technology for skin cancer diagnostics. However, the interobserver agreement (IOA) of known image markers of keratinocyte carcinomas (KC), including basal cell carcinoma (BCC) and squamous cell carcinoma (SCC), as well as precursors, SCC in situ (CIS) and actinic keratosis (AK), remains unexplored. This study determined IOA on the presence or absence of 10 key LC-OCT image markers of KC and precursors, among evaluators new to LC-OCT with different levels of dermatologic imaging experience. Secondly, the frequency and association between reported image markers and lesion types, was determined. Six evaluators blinded to histopathologic diagnoses, assessed 75 LC-OCT images of KC (21 SCC; 21 BCC), CIS (12), and AK (21). For each image, evaluators independently reported the presence or absence of 10 predefined key image markers of KCs and precursors described in an LC-OCT literature review. Evaluators were stratified by experience-level as experienced (3) or novices (3) based on previous OCT and reflectance confocal microscopy usage. IOA was tested for all groups, using Conger's kappa coefficient (κ). The frequency of reported image marker and their association with lesion-types, were calculated as proportions and odds ratios (OR), respectively. Overall IOA was highest for the image markers lobules (κ = 0.68, 95% confidence interval (CI) 0.57;0.78) and clefting (κ = 0.63, CI 0.52;0.74), typically seen in BCC (94%;OR 143.2 and 158.7, respectively, p < 0.001), followed by severe dysplasia (κ = 0.42, CI 0.31;0.53), observed primarily in CIS (79%;OR 7.1, p < 0.001). The remaining seven image-markers had lower IOA (κ = 0.06-0.32) and were more evenly observed across lesion types. The lowest IOA was noted for a well-defined (κ = 0.07, CI 0;0.15) and interrupted dermal-epidermal junction (DEJ) (κ = 0.06, CI -0.002;0.13). IOA was higher for all image markers among experienced evaluators versus novices. This study shows varying IOA for 10 key image markers of KC and precursors in LC-OCT images among evaluators new to the technology. IOA was highest for the assessments of lobules, clefting, and severe dysplasia while lowest for the assessment of the DEJ integrity.

Efficacy of Magnifying Endoscopy with Narrow-Band Imaging in the Diagnosis of Early Gastric Cancer and Gastric Intraepithelial Neoplasia.

Early diagnosis of gastric cancer can improve the prognosis of patients, especially for those with early gastric cancer (EGC), but only 15% of patients, or less, are diagnosed with EGC and precancerous lesions. Magnifying endoscopy with narrow-band imaging (ME-NBI) can improve diagnostic accuracy. We assess the efficacy of ME-NBI in diagnosing ECG and precancerous lesions, especially some characteristics under NBI+ME. This was a retrospective analysis of 131 patients with EGC or gastric intraepithelial neoplasia (IN) who had undergone endoscopic submucosal dissection and were pathologically diagnosed with EGC or IN according to 2019 WHO criteria for gastrointestinal tract tumors. We studied the characteristics of lesions under ME-NBI ,compared the diagnostic efficacy of ME-NBI and white light endoscopy (WLI) plus biopsy, and investigated the effect of Helicobacter pylori infection on microvascular and microsurface pattern. The diagnostic accuracy of ME-NBI for EGC, high-grade IN (HGIN), and low-grade IN (LGIN) was 76.06%, 77.96%, and 77.06%, respectively. The accuracy of WLI plus biopsy in diagnosing the above lesions was 69.7%, 57.5%, and 60.53%, respectively. The rate of gyrus-like tubular pattern was highest in LGIN (60.46%), whereas the highest rate of papillary pattern was 57.14% in HGIN and villous tubular pattern was 52% in EGC. Demarcation lines have better sensitivity for differentiating EGC from IN (92.06%). The ME-NBI has higher diagnostic accuracy for EGC than WLI plus biopsy. Demarcation lines and villous and papillary-like microsurface patterns are more specific as EGC and HGIN characteristics. The cerebral gyrus-like microsurface pattern is more specific for LGIN.

Surgical treatment of breast precancers - our experience.

INTRODUCTION: Thanks to mammographic screening and the improvement of breast cancer diagnostics, the detection of precancers is also increasing. They are defined as morphological changes of the mammary gland which are more likely to cause cancer. The evaluated precancers are atypical ductal hyperplasia (ADH), lobular carcinoma in situ (LCIS) and radial scar. METHODOLOGY: In the period 1. 1. 2018-31. 12. 2022, we performed 1,302 planned operations for breast disease at the Surgical Clinic of Teaching Hospital Plzen, of which 30 (2%) were precancer operations. ADH was confirmed 11×, LCIS 8×, and a radical scar 11×. The average age of the patients in all three groups was 56 years (27-85). Precancer was diagnosed 8× only by sonography, 3× by mammography and 19× by a combination of both methods. Subsequently, a puncture biopsy was always completed. We performed 28 tumor excisions with intraoperative biopsy and 2 mastectomies. RESULTS: In the case of ADH from puncture biopsy, ADH was confirmed intraoperatively 8×, DCIS was diagnosed 2×, and mucinous carcinoma 1×. In LCIS, no tumor was found by intraoperative biopsy 4×, LCIS was confirmed 1×, lobular invasive carcinoma was diagnosed 1×, mastectomy was performed 2× without intraoperative biopsy. In the radial scar, ADH was diagnosed 3×, sclerosing adenosis 6×, DCIS 1×, invasive carcinoma 1×. After the final histological processing of the samples, there was an increase in diagnosed carcinomas. In ADH, DCIS was confirmed 3×, DIC 2×, and mucinous carcinoma 1×. In LCIS, LIC was diagnosed 3×. In the radial scar, DCIS was confirmed 1×, and invasive carcinoma remain 1×. Thus, carcinoma was diagnosed in 11 patients (37%) thanks to the surgical solution. No patient underwent axillary node surgery. All 11 patients subsequently underwent oncological treatment, always a combination of radiotherapy and hormone therapy. All patients are alive, 10 patients are in complete remission of the disease, one with DCIS experienced a local recurrence after 4 years. CONCLUSION: Surgical treatment of precancers of the breast makes sense, DCIS or even invasive cancer is often hidden in addition to precancer. Thanks to the surgical solution, the cancer was detected in time.

A comprehensive dataset of annotated oral cavity images for diagnosis of oral cancer and oral potentially malignant disorders.

OBJECTIVES: This study aims to address the critical gap of unavailability of publicly accessible oral cavity image datasets for developing machine learning (ML) and artificial intelligence (AI) technologies for the diagnosis and prognosis of oral cancer (OCA) and oral potentially malignant disorders (OPMD), with a particular focus on the high prevalence and delayed diagnosis in Asia. MATERIALS AND METHODS: Following ethical approval and informed written consent, images of the oral cavity were obtained from mobile phone cameras and clinical data was extracted from hospital records from patients attending to the Dental Teaching Hospital, Peradeniya, Sri Lanka. After data management and hosting, image categorization and annotations were done by clinicians using a custom-made software tool developed by the research team. RESULTS: A dataset comprising 3000 high-quality, anonymized images obtained from 714 patients were classified into four distinct categories: healthy, benign, OPMD, and OCA. Images were annotated with polygonal shaped oral cavity and lesion boundaries. Each image is accompanied by patient metadata, including age, sex, diagnosis, and risk factor profiles such as smoking, alcohol, and betel chewing habits. CONCLUSION: Researchers can utilize the annotated images in the COCO format, along with the patients' metadata, to enhance ML and AI algorithm development.

Multi-Instance Learning for Vocal Fold Leukoplakia Diagnosis Using White Light and Narrow-Band Imaging: A Multicenter Study.

OBJECTIVES: Vocal fold leukoplakia (VFL) is a precancerous lesion of laryngeal cancer, and its endoscopic diagnosis poses challenges. We aim to develop an artificial intelligence (AI) model using white light imaging (WLI) and narrow-band imaging (NBI) to distinguish benign from malignant VFL. METHODS: A total of 7057 images from 426 patients were used for model development and internal validation. Additionally, 1617 images from two other hospitals were used for model external validation. Modeling learning based on WLI and NBI modalities was conducted using deep learning combined with a multi-instance learning approach (MIL). Furthermore, 50 prospectively collected videos were used to evaluate real-time model performance. A human-machine comparison involving 100 patients and 12 laryngologists assessed the real-world effectiveness of the model. RESULTS: The model achieved the highest area under the receiver operating characteristic curve (AUC) values of 0.868 and 0.884 in the internal and external validation sets, respectively. AUC in the video validation set was 0.825 (95% CI: 0.704-0.946). In the human-machine comparison, AI significantly improved AUC and accuracy for all laryngologists (p < 0.05). With the assistance of AI, the diagnostic abilities and consistency of all laryngologists improved. CONCLUSIONS: Our multicenter study developed an effective AI model using MIL and fusion of WLI and NBI images for VFL diagnosis, particularly aiding junior laryngologists. However, further optimization and validation are necessary to fully assess its potential impact in clinical settings. LEVEL OF EVIDENCE: 3 Laryngoscope, 134:4321-4328, 2024.

Clinical significance of 18F-FDG-PET/CT for detection of incidental pre-malignant and malignant colonic lesions: correlation with colonoscopic and histopathological results.

BACKGROUND: Incidental colorectal fluorodeoxyglucose (FDG) uptake, observed during positron emission tomography/computed tomography (PET/CT) scans, attracts particular attention due to its potential to represent both benign and pre-malignant/malignant lesions. Early detection and excision of these lesions are crucial for preventing cancer development and reducing mortality. This research aims to evaluate the correlation between incidental colorectal FDG uptake on PET/CT with colonoscopic and histopathological results. METHODS: Retrospective analysis was performed on data from all patients who underwent PET/CT between December 2019 and December 2023 in our hospital. The study included 79 patients with incidental colonic FDG uptake who underwent endoscopy. Patient characteristics, imaging parameters, and the corresponding colonoscopy and histopathological results were studied. A comparative analysis was performed among the findings from each of these modalities. The optimal cut-off value of SUVmax for 18F-FDG PET/CT diagnosis of premalignant and malignant lesions was determined by receiver operating characteristic (ROC) curves. The area under the curve (AUC) of SUVmax and the combined parameters of SUVmax and colonic wall thickening (CWT) were analyzed. RESULTS: Among the 79 patients with incidental colorectal FDG uptake, histopathology revealed malignancy in 22 (27.9%) patients and premalignant polyps in 22 (27.9%) patients. Compared to patients with benign lesions, patients with premalignant and malignant lesions were more likely to undergo a PET/CT scan for primary evaluation (p = 0.013), and more likely to have focal GIT uptake (p = 0.001) and CWT (p = 0.001). A ROC curve analysis was made and assesed a cut-off value of 7.66 SUVmax (sensitivity: 64.9% and specificity: 82.4%) to distinguish premalignant and malignant lesions from benign lesions. The AUCs of the SUVmax and the combined parameters of SUVmax and CWT were 0.758 and 0.832 respectively. CONCLUSION: For patients undergo PET/CT for primary evaluation, imaging features of colorectal focal FDG uptake and CWT were more closely associated with premalignant and malignant lesions. The SUVmax helps determine benign and premalignant/malignant lesions of the colorectum. Moreover, the combination of SUVmax and CWT parameters have higher accuracy in estimating premalignant and malignant lesions than SUVmax.

Artificial intelligence for gastric cancer in endoscopy: From diagnostic reasoning to market.

Recognition of gastric conditions during endoscopy exams, including gastric cancer, usually requires specialized training and a long learning curve. Besides that, the interobserver variability is frequently high due to the different morphological characteristics of the lesions and grades of mucosal inflammation. In this sense, artificial intelligence tools based on deep learning models have been developed to support physicians to detect, classify, and predict gastric lesions more efficiently. Even though a growing number of studies exists in the literature, there are multiple challenges to bring a model to practice in this field, such as the need for more robust validation studies and regulatory hurdles. Therefore, the aim of this review is to provide a comprehensive assessment of the current use of artificial intelligence applied to endoscopic imaging to evaluate gastric precancerous and cancerous lesions and the barriers to widespread implementation of this technology in clinical routine.

Precancerous Change Detection Technique on Mammography Breast Cancer Images based on Mean Ratio and Log Ratio using Fuzzy c Mean Classification with Gabor Filter

BACKGROUND: The growing rate of breast cancer necessitates immediate global attention. Mammography images are used to determine the stage of malignancy. Breast cancer stages must be identified in order to save a person's life. OBJECTIVE: This article's main goal is to identify different techniques to obtain the difference between two breast cancer mammography images taken of the same individual at different times. This is the first effort to identify breast cancer in mammography images using change detection techniques. The Mammogram Image Change Detection (ICD) technique is also a recent advancement to prevent breast cancer in the early stage and precancerous level in medical images. METHODS: The main purpose of this work is to observe the changes between breast cancer images in different screening periods using different techniques. Mammogram Breast Cancer Image Change Detection (MBCICD) methods usually start with a Difference Image (DI) and classify the pixels in the DI into changed and unaffected classes using unsupervised fuzzy c means (FCM) clustering methods based on texture features taken from the log and mean ratio difference pictures. Two operators, mean ratio and log ratio, were used to check the changes in the images. The Gabor wavelet is utilized as a feature extraction technique among several standards. Using the Gabor wavelet ratio operators is a useful method for altering the detection of breast cancer in mammography images. Currently, it is challenging to obtain real malignant images of the same person for testing or training. In this study, two images are utilized. To clearly see the changes, one is an image from the MIAS breast cancer mammography images dataset, and the other is a self-generated change image. RESULTS: The research aims to examine the image results and other quantitative analysis results of proposed change detection methods on cancer images. The Mean Ratio Accuracy result is 0.9738, and the Log ratio PCC is 0.9737. The classification results are the Log Ratio + Gabor Filter + FCM is 0.9737, and Mean Ratio +Gabor Filter + FCM is 0.9719. The mean Ratio Accuracy result is 0.9738, Log ratio is 0.9737. Log Ratio + Gabor Filter + FCM is 0.9737, Mean Ratio +Gabor Filter + FCM is 0.9719. Comparing the PCC of proposed change detection methods with the FDA-RMG method on the same dataset, the accuracy is 0.9481 only. CONCLUSION: The study concludes that variations in mammography breast cancer images could be successfully identified using the ratio operators with Gabor wavelet features.

Early features of pancreatic cancer on magnetic resonance imaging (MRI): a case-control study.

PURPOSE: Magnetic resonance imaging has been recommended as a primary imaging modality among high-risk individuals undergoing screening for pancreatic cancer. We aimed to delineate potential precursor lesions for pancreatic cancer on MR imaging. METHODS: We conducted a case-control study at Kaiser Permanente Southern California (2008-2018) among patients that developed pancreatic cancer who had pre-diagnostic MRI examinations obtained 2-36 months prior to cancer diagnosis (cases) matched 1:2 by age, gender, race/ethnicity, contrast status and year of scan (controls). Patients with history of acute/chronic pancreatitis or prior pancreatic surgery were excluded. Images underwent blind review with assessment of a priori defined series of parenchymal and ductal features. We performed logistic regression to assess the associations between individual factors and pancreatic cancer. We further assessed the interaction among features as well as performed a sensitivity analysis stratifying based on specific time-windows (2-3 months, 4-12 months, 13-36 months prior to cancer diagnosis). RESULTS: We identified 141 cases (37.9% stage I-II, 2.1% III, 31.4% IV, 28.6% unknown) and 292 matched controls. A solid mass was noted in 24 (17%) of the pre-diagnostic MRI scans. Compared to controls, pre-diagnostic images from cancer cases more frequently exhibited the following ductal findings: main duct dilatation (51.4% vs 14.3%, OR [95% CI]: 7.75 [4.19-15.44], focal pancreatic duct stricture with distal (upstream) dilatation (43.6% vs 5.6%, OR 12.71 [6.02-30.89], irregularity (42.1% vs 6.0%, OR 9.73 [4.91-21.43]), focal pancreatic side branch dilation (13.6% vs1.6%, OR 11.57 [3.38-61.32]) as well as parenchymal features: atrophy (57.9% vs 27.4%, OR 46.4 [2.71-8.28], focal area of signal abnormality (39.3% vs 4.8%, OR 15.69 [6.72-44,78]), all p < 0.001). CONCLUSION: In addition to potential missed lesions, we have identified a series of ductal and parenchymal features on MRI that are associated with increased odds of developing pancreatic cancer.

Multiscale Optical Imaging Fusion for Cervical Precancer Diagnosis: Integrating Widefield Colposcopy and High-Resolution Endomicroscopy.

OBJECTIVE: Early detection and treatment of cervical precancers can prevent disease progression. However, in low-resource communities with a high incidence of cervical cancer, high equipment costs and a shortage of specialists hinder preventative strategies. This manuscript presents a low-cost multiscale in vivo optical imaging system coupled with a computer-aided diagnostic system that could enable accurate, real-time diagnosis of high-grade cervical precancers. METHODS: The system combines portable colposcopy and high-resolution endomicroscopy (HRME) to acquire spatially registered widefield and microscopy videos. A multiscale imaging fusion network (MSFN) was developed to identify cervical intraepithelial neoplasia grade 2 or more severe (CIN 2+). The MSFN automatically identifies and segments the ectocervix and lesions from colposcopy images, extracts nuclear morphology features from HRME videos, and integrates the colposcopy and HRME information. RESULTS: With a threshold value set to achieve sensitivity equal to clinical impression (0.98 [p = 1.0]), the MSFN achieved a significantly higher specificity than clinical impression (0.75 vs. 0.43, p = 0.000006). CONCLUSION: Our findings show that multiscale optical imaging of the cervix allows the highly sensitive and specific detection of high-grade precancers. SIGNIFICANCE: The multiscale imaging system and MSFN could facilitate the accurate, real-time diagnosis of cervical precancers in low-resource settings.

The insufficiency of CT examination in early detection of central lung squamous cell carcinoma and squamous epithelial precancerous lesions.

BACKGROUND: CT examination for lung cancer has been carried out for more than 20 years and great achievements have been made in the early detection of lung cancer. However, in the clinical work, a large number of advanced central lung squamous cell carcinoma are still detected through bronchoscopy. Meanwhile, a part of CT-occult central lung squamous cell carcinoma and squamous epithelial precancerous lesions are also accidentally detected through bronchoscopy. METHODS: This study retrospectively collects the medical records of patients in the bronchoscopy room of the Endoscopy Department of Zhejiang Cancer Hospital from January 2014 to December 2018. The inclusion criteria for patients includes: 1.Patient medical records completed, 2.Without history of lung cancer before the diagnosis and first pathological diagnosis of primary lung cancer, 3.Have the lung CT data of the same period, 4.Have the bronchoscopy records and related pathological diagnosis, 5.The patients undergoing radical surgical treatment must have a complete postoperative pathological diagnosis. Finally, a total of 10,851 patients with primary lung cancer are included in the study, including 7175 males and 3676 females, aged 22-98 years. Firstly, 130 patients with CT-occult lesions are extracted and their clinical features are analyzed. Then, 604 cases of single central squamous cell carcinoma and 3569 cases of peripheral adenocarcinoma are extracted and compares in postoperative tumor diameter and lymph node metastasis. RESULTS: 115 cases of CT-occult central lung squamous cell carcinoma and 15 cases of squamous epithelial precancerous lesions are found. In the total lung cancer, the proportion of CT-occult lesions is 130/10,851 (1.20%). Meanwhile, all these patients are middle-aged and elderly men with a history of heavy smoking. There are statistically significant differences in postoperative median tumor diameter (3.65 cm vs.1.70 cm, P < 0.0001) and lymph node metastasis rate (50.99% vs.13.06%, P < 0.0001) between 604 patients with operable single central lung squamous cell carcinoma and 3569 patients with operable peripheral lung adenocarcinoma. Of the 604 patients with squamous cell carcinoma, 96.52% (583/604) are male with a history of heavy smoking and aged 40-82 years with a median age of 64 years. CONCLUSIONS: This study indicates that the current lung CT examination of lung cancer is indeed insufficiency for the early diagnosis of central squamous cell carcinoma and squamous epithelial precancerous lesions. Further bronchoscopy in middle-aged and elderly men with a history of heavy smoking can make up for the lack of routine lung CT examination.

A deep learning model based on magnifying endoscopy with narrow-band imaging to evaluate intestinal metaplasia grading and OLGIM staging: A multicenter study.

BACKGROUND AND PURPOSE: Patients with stage III or IV of operative link for gastric intestinal metaplasia assessment (OLGIM) are at a higher risk of gastric cancer (GC). We aimed to construct a deep learning (DL) model based on magnifying endoscopy with narrow-band imaging (ME-NBI) to evaluate OLGIM staging. METHODS: This study included 4473 ME-NBI images obtained from 803 patients at three endoscopy centres. The endoscopic expert marked intestinal metaplasia (IM) regions on endoscopic images of the target biopsy sites. Faster Region-Convolutional Neural Network model was used to grade IM lesions and predict OLGIM staging. RESULTS: The diagnostic performance of the model for IM grading in internal and external validation sets, as measured by the area under the curve (AUC), was 0.872 and 0.803, respectively. The accuracy of this model in predicting the high-risk stage of OLGIM was 84.0%, which was not statistically different from that of three junior (71.3%, p = 0.148) and three senior endoscopists (75.3%, p = 0.317) specially trained in endoscopic images corresponding to pathological IM grade, but higher than that of three untrained junior endoscopists (64.0%, p = 0.023). CONCLUSION: This DL model can assist endoscopists in predicting OLGIM staging using ME-NBI without biopsy, thereby facilitating screening high-risk patients for GC.

Narrow band imaging reveals field cancerisation undetected by conventional white light: Optical diagnosis versus histopathology.

OBJECTIVE: To assess whether narrow band imaging (NBI) detects fields of cancerisation around suspicious lesions in the upper aerodigestive tract, which were undetected by white light imaging (WLI). METHODS: In 96 patients with laryngeal and pharyngeal lesions suspicious for malignancy, 206 biopsies were taken during laryngoscopy: 96 biopsies of suspicious lesions detected by both WLI and NBI (WLI+/NBI+), 60 biopsies adjacent mucosa only suspicious with NBI (WLI-/NBI+), and 46 biopsies of NBI and WLI unsuspicious mucosa (WLI-/NBI-) as negative controls. Optical diagnosis according to the Ni-classification was compared with histopathology. RESULTS: Signs of (pre)malignancy were found in 88% of WLI+/NBI+ biopsies, 32% of WLI-/NBI+ biopsies and 0% in WLI-/NBI- (p < .001). In 58% of the WLI-/NBI+ mucosa any form of dysplasia or carcinoma was detected. CONCLUSION: The use of additional NBI led to the detection of (pre)malignancy in 32% of the cases, that would have otherwise remained undetected with WLI alone. This highlights the potential of NBI as a valuable adjunct to WLI in the identification of suspicious lesions in the upper aerodigestive tract.

OralEpitheliumDB: A Dataset for Oral Epithelial Dysplasia Image Segmentation and Classification.

Early diagnosis of potentially malignant disorders, such as oral epithelial dysplasia, is the most reliable way to prevent oral cancer. Computational algorithms have been used as an auxiliary tool to aid specialists in this process. Usually, experiments are performed on private data, making it difficult to reproduce the results. There are several public datasets of histological images, but studies focused on oral dysplasia images use inaccessible datasets. This prevents the improvement of algorithms aimed at this lesion. This study introduces an annotated public dataset of oral epithelial dysplasia tissue images. The dataset includes 456 images acquired from 30 mouse tongues. The images were categorized among the lesion grades, with nuclear structures manually marked by a trained specialist and validated by a pathologist. Also, experiments were carried out in order to illustrate the potential of the proposed dataset in classification and segmentation processes commonly explored in the literature. Convolutional neural network (CNN) models for semantic and instance segmentation were employed on the images, which were pre-processed with stain normalization methods. Then, the segmented and non-segmented images were classified with CNN architectures and machine learning algorithms. The data obtained through these processes is available in the dataset. The segmentation stage showed the F1-score value of 0.83, obtained with the U-Net model using the ResNet-50 as a backbone. At the classification stage, the most expressive result was achieved with the Random Forest method, with an accuracy value of 94.22%. The results show that the segmentation contributed to the classification results, but studies are needed for the improvement of these stages of automated diagnosis. The original, gold standard, normalized, and segmented images are publicly available and may be used for the improvement of clinical applications of CAD methods on oral epithelial dysplasia tissue images.

Computer-aided diagnosis in real-time endoscopy for all stages of gastric carcinogenesis: Development and validation study.

OBJECTIVE: Using endoscopic images, we have previously developed computer-aided diagnosis models to predict the histopathology of gastric neoplasms. However, no model that categorizes every stage of gastric carcinogenesis has been published. In this study, a deep-learning-based diagnosis model was developed and validated to automatically classify all stages of gastric carcinogenesis, including atrophy and intestinal metaplasia, in endoscopy images. DESIGN: A total of 18,701 endoscopic images were collected retrospectively and randomly divided into train, validation, and internal-test datasets in an 8:1:1 ratio. The primary outcome was lesion-classification accuracy in six categories: normal/atrophy/intestinal metaplasia/dysplasia/early /advanced gastric cancer. External-validation of performance in the established model used 1427 novel images from other institutions that were not used in training, validation, or internal-tests. RESULTS: The internal-test lesion-classification accuracy was 91.2% (95% confidence interval: 89.9%-92.5%). For performance validation, the established model achieved an accuracy of 82.3% (80.3%-84.3%). The external-test per-class receiver operating characteristic in the diagnosis of atrophy and intestinal metaplasia was 93.4 ± 0% and 91.3 ± 0%, respectively. CONCLUSIONS: The established model demonstrated high performance in the diagnosis of preneoplastic lesions (atrophy and intestinal metaplasia) as well as gastric neoplasms.

CrossU-Net: Dual-modality cross-attention U-Net for segmentation of precancerous lesions in gastric cancer.

Gastric precancerous lesions (GPL) significantly elevate the risk of gastric cancer, and precise diagnosis and timely intervention are critical for patient survival. Due to the elusive pathological features of precancerous lesions, the early detection rate is less than 10%, which hinders lesion localization and diagnosis. In this paper, we provide a GPL pathological dataset and propose a novel method for improving the segmentation accuracy on a limited-scale dataset, namely RGB and Hyperspectral dual-modal pathological image Cross-attention U-Net (CrossU-Net). Specifically, we present a self-supervised pre-training model for hyperspectral images to serve downstream segmentation tasks. Secondly, we design a dual-stream U-Net-based network to extract features from different modal images. To promote information exchange between spatial information in RGB images and spectral information in hyperspectral images, we customize the cross-attention mechanism between the two networks. Furthermore, we use an intermediate agent in this mechanism to improve computational efficiency. Finally, we add a distillation loss to align predicted results for both branches, improving network generalization. Experimental results show that our CrossU-Net achieves accuracy and Dice of 96.53% and 91.62%, respectively, for GPL lesion segmentation, providing a promising spectral research approach for the localization and subsequent quantitative analysis of pathological features in early diagnosis.

Deep Learning in High-Resolution Anoscopy: Assessing the Impact of Staining and Therapeutic Manipulation on Automated Detection of Anal Cancer Precursors.

INTRODUCTION: High-resolution anoscopy (HRA) is the gold standard for detecting anal squamous cell carcinoma (ASCC) precursors. Preliminary studies on the application of artificial intelligence (AI) models to this modality have revealed promising results. However, the impact of staining techniques and anal manipulation on the effectiveness of these algorithms has not been evaluated. We aimed to develop a deep learning system for automatic differentiation of high-grade squamous intraepithelial lesion vs low-grade squamous intraepithelial lesion in HRA images in different subsets of patients (nonstained, acetic acid, lugol, and after manipulation). METHODS: A convolutional neural network was developed to detect and differentiate high-grade and low-grade anal squamous intraepithelial lesions based on 27,770 images from 103 HRA examinations performed in 88 patients. Subanalyses were performed to evaluate the algorithm's performance in subsets of images without staining, acetic acid, lugol, and after manipulation of the anal canal. The sensitivity, specificity, accuracy, positive and negative predictive values, and area under the curve were calculated. RESULTS: The convolutional neural network achieved an overall accuracy of 98.3%. The algorithm had a sensitivity and specificity of 97.4% and 99.2%, respectively. The accuracy of the algorithm for differentiating high-grade squamous intraepithelial lesion vs low-grade squamous intraepithelial lesion varied between 91.5% (postmanipulation) and 100% (lugol) for the categories at subanalysis. The area under the curve ranged between 0.95 and 1.00. DISCUSSION: The introduction of AI to HRA may provide an accurate detection and differentiation of ASCC precursors. Our algorithm showed excellent performance at different staining settings. This is extremely important because real-time AI models during HRA examinations can help guide local treatment or detect relapsing disease.

Artificial intelligence-aided diagnostic imaging: A state-of-the-art technique in precancerous screening.

BACKGROUND AND AIM: Chromoendoscopy with the use of indigo carmine (IC) dye is a crucial endoscopic technique to identify gastrointestinal neoplasms. However, its performance is limited by the endoscopist's skill, and no standards are available for lesion identification. Thus, we developed an artificial intelligence (AI) model to replace chromoendoscopy. METHODS: This pilot study assessed the feasibility of our novel AI model in the conversion of white-light images (WLI) into virtual IC-dyed images based on a generative adversarial network. The predictions of our AI model were evaluated against the assessments of five endoscopic experts who were blinded to the purpose of this study with a staining quality rating from 1 (unacceptable) to 4 (excellent). RESULTS: The AI model successfully transformed the WLI of polyps with different morphologies and different types of lesions in the gastrointestinal tract into virtual IC-dyed images. The quality ratings of the real IC-dyed and AI images did not significantly differ concerning surface structure (AI vs IC: 3.08 vs 3.00), lesion border (3.04 vs 2.98), and overall contrast (3.14 vs 3.02) from 10 sets of images (10 AI images and 10 real IC-dyed images). Although the score depended significantly on the evaluator, the staining methods (AI or real IC) and evaluators had no significant interaction (P > 0.05) with each other. CONCLUSION: Our results demonstrated the feasibility of employing AI model's virtual IC staining, increasing the possibility of being employed in daily practice. This novel technology may facilitate gastrointestinal lesion identification in the future.