Classification Of Lesions Research Articles

Automated segmentation for early detection of uveal melanoma

ObjectifLe mélanome de l'uvée est le cancer intraoculaire le plus fréquent chez l'adulte. À l'heure actuelle, les méthodes de dépistage et de triage des tumeurs choroïdiennes mélanotiques se butent à des obstacles intrinsèques, surtout dans les régions mal desservies par des oncologues oculaires spécialisés. La présente étude se penche sur le potentiel de l'apprentissage artificiel pour automatiser la segmentation tumorale. Nous avons créé et évalué un modèle d'apprentissage artificiel qui repose sur des photographies ultra-grand champ du fond d’œil pour réaliser la segmentation des lésions. MéthodeOn a procédé à un examen des dossiers médicaux de patients qui ont reçu un diagnostic de mélanome de l'uvée, de naevus choroïdien ou d'hypertrophie congénitale de l’épithélium pigmentaire rétinien dans un centre universitaire de soins tertiaires. On a réalisé une seule photographie ultra-grand champ (Optos PLC, Dunfermline, Fife, Écosse) du fond d’œil des patients retenus; la photographie était de qualité suffisante pour visualiser les lésions d'intérêt, comme l'a confirmé un oncologue oculaire unique. Ces images ont servi à créer et à tester un algorithme d'apprentissage artificiel en vue de la segmentation de lésions. RésultatsAu total, 396 images ont servi à créer un algorithme d'apprentissage artificiel en vue de la segmentation de lésions, et 90 images supplémentaires ont servi d'ensemble de données à tester parallèlement à des images de 30 sujets témoins en bonne santé. Parmi les images sur lesquelles le modèle d'apprentissage a réussi à identifier des lésions, la segmentation automatisée a donné un coefficient de similarité de Dice de 0,86, de 0,81 et de 0,85 pour le mélanome de l'uvée, le naevus choroïdien et l'hypertrophie congénitale de l’épithélium pigmentaire rétinien, respectivement. La sensibilité envers toute lésion décelée sur une image donnée se chiffrait à 1,00, à 0,90 et à 0,87, respectivement. La spécificité était de 0,93 dans les images sans lésion. ConclusionNotre étude a permis d'objectiver la capacité d'un nouvel algorithme d'apprentissage artificiel et d’évoquer son utilité clinique potentielle et sa grande accessibilité en tant que méthode de dépistage des tumeurs choroïdiennes. D'autres méthodes d’évaluation s'imposent pour améliorer davantage la précision du modèle en matière de classification des lésions et de détermination du diagnostic.

Endoscopic evaluation of gastric mucosal alterations in patients with chronic heart failure II-III nyha functional class utilizing advanced endoscopic techniques

Objective: To assess the alterations of gastric mucosa, its pathomorphological and functional characteristics in patients with complete erosions suffering from chronic heart failure II-III NYHA functional class. Materials and Methods: 185 patients with complete erosions comprised the main group, while the control group consisted of 183 patients with incomplete erosions of the stomach. Patients in both groups had chronic heart failure II-III NYHA functional class. All patients underwent esophagogastroduodenoscopy with white light imaging (WLI), narrow-band imaging (NBI), and ZOOM modes, as well as biopsy sampling followed by pathomorphological examination, determination of Helicobacter pylori presence, and endoscopic pH-metry. The diagnosis of chronic heart failure was established based on anamnestic data and the results of instrumental and functional diagnostic methods, including electrocardiography (ECG), echocardiography, Holter monitoring of ECG, stress echocardiography, and coronary angiography. Results: The study utilized the classification of gastric erosive lesions by K. Toljamo (2012), according to which in the main group, mature erosions subtype Ia were detected in 83.8% of cases, while immature extensive erosions subtype Ib were found in 16.2%. In combination with haemorrhagic erosions (type III), complete erosions were detected in 12.6% of cases, while with incomplete erosions of type II, they were found in 11.4%. The combination of complete erosions with erosive prepyloric changes (type IV) was identified in 8.7%. In the control group, gastric erosive lesions were mainly represented by type III (41%) and type IV (38.8%). Based on the analysis of endoscopic changes in the gastric mucosa using WLI, NBI, and ZOOM modes, the most characteristic visual changes were identified for each type and subtype of erosive gastric mucosa alterations. In patients with chronic heart failure II-III NYHA functional class, gastritis was noted in 75.1% of cases in the main group and in 58.5% in the control group. Helicobacter pylori infection of weak degree was determined in 40.5% in the main group and 37.7% in the control group. Evaluation of endoscopic pH-metry showed predominantly hypochlorhydria in both groups - 70.8% in the main group and 48.6% in the control group. Conclusion: By comparing the macroscopic features of various gastric erosive lesions in WLI, NBI, and ZOOM modes according to the classification by K. Toljamo (2012), the most characteristic signs for each type and subtype were identified. High prevalence of gastritis was established among patients with chronic heart failure II-III NYHA functional class.

A hybrid CNN with transfer learning for skin cancer disease detection.

The leading cause of cancer-related deaths worldwide is skin cancer. Effective therapy depends on the early diagnosis of skin cancer through the precise classification of skin lesions. However, dermatologists may find it difficult and time-consuming to accurately classify skin lesions. The use of transfer learning to boost skin cancer classification model precision is a promising strategy. In this work, we proposed a hybrid CNN with a transfer learning model and a random forest classifier for skin cancer disease detection. To evaluate the efficacy of the proposed model, it was verified over two datasets of benign skin moles and malignant skin moles. The proposed model is able to classify images with an accuracy of up to 90.11%. The empirical results and analysis assure the feasibility and effectiveness of the proposed model for skin cancer classification.

Robust prostate disease classification using transformers with discrete representations

Purpose:Automated prostate disease classification on multi-parametric MRI has recently shown promising results with the use of convolutional neural networks (CNNs). The vision transformer (ViT) is a convolutional free architecture which only exploits the self-attention mechanism and has surpassed CNNs in some natural imaging classification tasks. However, these models are not very robust to textural shifts in the input space. In MRI, we often have to deal with textural shift arising from varying acquisition protocols. Here, we focus on the ability of models to generalise well to new magnet strengths for MRI.Method:We propose a new framework to improve the robustness of vision transformer-based models for disease classification by constructing discrete representations of the data using vector quantisation. We sample a subset of the discrete representations to form the input into a transformer-based model. We use cross-attention in our transformer model to combine the discrete representations of T2-weighted and apparent diffusion coefficient (ADC) images.Results:We analyse the robustness of our model by training on a 1.5 T scanner and test on a 3 T scanner and vice versa. Our approach achieves SOTA performance for classification of lesions on prostate MRI and outperforms various other CNN and transformer-based models in terms of robustness to domain shift and perturbations in the input space.Conclusion:We develop a method to improve the robustness of transformer-based disease classification of prostate lesions on MRI using discrete representations of the T2-weighted and ADC images.

In Vivo Classification and Characterization of Carotid Atherosclerotic Lesions with Integrated 18F-FDG PET/MRI.

The aim of this study was to exploit integrated PET/MRI to simultaneously evaluate the morphological, component, and metabolic features of advanced atherosclerotic plaques and explore their incremental value. In this observational prospective cohort study, patients with advanced plaque in the carotid artery underwent 18F-FDG PET/MRI. Plaque morphological features were measured, and plaque component features were determined via MRI according to AHA lesion-types. Maximum standardized uptake values (SUVmax) and tissue to background ratio (TBR) on PET were calculated. Area under the receiver-operating characteristic curve (AUC) and net reclassification improvement (NRI) were used to compare the incremental contribution of FDG uptake when added to AHA lesion-types for symptomatic plaque classification. A total of 280 patients with advanced plaque in the carotid artery were recruited. A total of 402 plaques were confirmed, and 87 of 402 (21.6%) were symptomatic plaques. 18F-FDG PET/MRI was performed a mean of 38 days (range 1-90) after the symptom. Increased stenosis degree (61.5% vs. 50.0%, p < 0.001) and TBR (2.96 vs. 2.32, p < 0.001) were observed in symptomatic plaques compared with asymptomatic plaques. The performance of the combined model (AHA lesion type VI + stenosis degree + TBR) for predicting symptomatic plaques was the best among all models (AUC = 0.789). The improvement of the combined model (AHA lesion type VII + stenosis degree + TBR) over AHA lesion type VII model for predicting symptomatic plaques was the highest (AUC = 0.757/0.454, combined model/AHA lesion type VII model), and the NRI was 50.7%. Integrated PET/MRI could simultaneously evaluate the morphological component and inflammation features of advanced atherosclerotic plaques and provide supplementary optimization information over AHA lesion-types for identifying vulnerable plaques in atherosclerosis subjects to achieve further stratification of stroke risk.

Analytic validation of convolutional neural network-based classification of pigmented skin lesions (PSLs) using unseen PSL hyperspectral data for clinical applications

Analytic validation of convolutional neural network-based classification of pigmented skin lesions (PSLs) using unseen PSL hyperspectral data for clinical applications

Classification of cervical lesions based on multimodal features fusion

Cervical cancer is a severe threat to women's health worldwide with a long cancerous cycle and a clear etiology, making early screening vital for the prevention and treatment. Based on the dataset provided by the Obstetrics and Gynecology Hospital of Fudan University, a four-category classification model for cervical lesions including Normal, low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL) and cancer (Ca) is developed. Considering the dataset characteristics, to fully utilize the research data and ensure the dataset size, the model inputs include original and acetic colposcopy images, lesion segmentation masks, human papillomavirus (HPV), thinprep cytologic test (TCT) and age, but exclude iodine images that have a significant overlap with lesions under acetic images. Firstly, the change information between original and acetic images is introduced by calculating the acetowhite opacity to mine the correlation between the acetowhite thickness and lesion grades. Secondly, the lesion segmentation masks are utilized to introduce prior knowledge of lesion location and shape into the classification model. Lastly, a cross-modal feature fusion module based on the self-attention mechanism is utilized to fuse image information with clinical text information, revealing the features correlation. Based on the dataset used in this study, the proposed model is comprehensively compared with five excellent models over the past three years, demonstrating that the proposed model has superior classification performance and a better balance between performance and complexity. The modules ablation experiments further prove that each proposed improved module can independently improve the model performance.

BRAIxDet: Learning to detect malignant breast lesion with incomplete annotations

Methods to detect malignant lesions from screening mammograms are usually trained with fully annotated datasets, where images are labelled with the localisation and classification of cancerous lesions. However, real-world screening mammogram datasets commonly have a subset that is fully annotated and another subset that is weakly annotated with just the global classification (i.e., without lesion localisation). Given the large size of such datasets, researchers usually face a dilemma with the weakly annotated subset: to not use it or to fully annotate it. The first option will reduce detection accuracy because it does not use the whole dataset, and the second option is too expensive given that the annotation needs to be done by expert radiologists. In this paper, we propose a middle-ground solution for the dilemma, which is to formulate the training as a weakly- and semi-supervised learning problem that we refer to as malignant breast lesion detection with incomplete annotations. To address this problem, our new method comprises two stages, namely: (1) pre-training a multi-view mammogram classifier with weak supervision from the whole dataset, and (2) extending the trained classifier to become a multi-view detector that is trained with semi-supervised student–teacher learning, where the training set contains fully and weakly-annotated mammograms. We provide extensive detection results on two real-world screening mammogram datasets containing incomplete annotations and show that our proposed approach achieves state-of-the-art results in the detection of malignant breast lesions with incomplete annotations.

SkinSwinViT: A Lightweight Transformer-Based Method for Multiclass Skin Lesion Classification with Enhanced Generalization Capabilities

In recent decades, skin cancer has emerged as a significant global health concern, demanding timely detection and effective therapeutic interventions. Automated image classification via computational algorithms holds substantial promise in significantly improving the efficacy of clinical diagnoses. This study is committed to mitigating the challenge of diagnostic accuracy in the classification of multiclass skin lesions. This endeavor is inherently formidable owing to the resemblances among various lesions and the constraints associated with extracting precise global and local image features within diverse dimensional spaces using conventional convolutional neural network methodologies. Consequently, this study introduces the SkinSwinViT methodology for skin lesion classification, a pioneering model grounded in the Swin Transformer framework featuring a global attention mechanism. Leveraging the inherent cross-window attention mechanism within the Swin Transformer architecture, the model adeptly captures local features and interdependencies within skin lesion images while additionally incorporating a global self-attention mechanism to discern overarching features and contextual information effectively. The evaluation of the model’s performance involved the ISIC2018 challenge dataset. Furthermore, data augmentation techniques augmented training dataset size and enhanced model performance. Experimental results highlight the superiority of the SkinSwinViT method, achieving notable metrics of accuracy, recall, precision, specificity, and F1 score at 97.88%, 97.55%, 97.83%, 99.36%, and 97.79%, respectively.

Comparative analysis of deep learning techniques for accurate stroke detection

Background: The traditional diagnosis of strokes through computed tomography (CT) heavily relies on radiologists’ expertise for accurate interpretation. However, the increasing demand for this critical task exceeds the available radiologist workforce, necessitating innovative solutions. This research addresses this challenge by introducing deep learning techniques to enhance the initial screening of stroke cases, thereby augmenting radiologists’ diagnostic capabilities. Objective: This study aims to compare four techniques for classifying stroke lesions in CT images. Materials and methods: Four distinct models-CNN-2-Model, LeNet, GoogleNet, and VGG-16-were trained using a dataset comprising 1,636 CT images, including 1,111 normal brain images and 525 stroke images. Seventy percent of the images were used to train the most effective deep learning model, and subsequently, these images were utilized to evaluate the performance of each model. The evaluation involved assessing accuracy, precision, sensitivity, specificity, F1 score, false positive rate, and AUC. Results: The evaluation process included a comprehensive statistical analysis of the models’ prediction results. The findings revealed that VGG-16 emerged as the top-performing deep learning model, achieving an impressive accuracy of 0.969, precision of 0.952, sensitivity of 0.952, specificity of 0.978, F1 score of 0.952, false positive rate of 0.022, and AUC of 0.965. Conclusion: In conclusion, deep learning techniques, particularly the VGG-16 model, demonstrate significant promise in enhancing the accuracy of stroke lesion classification in CT images. These findings underscore the potential of leveraging advanced technologies to address the growing challenges in stroke diagnosis and pave the way for more efficient and accessible healthcare solutions.

The value of elastography strain rate ratio in benign and malignant BI-RADS-US 3-4 breast masses.

In the realm of breast diagnostics, Breast Imaging Reporting and Data System (BI-RADS) serves as a systematic framework, guiding a methodical exploration into the nuanced narratives of masses. The current study aims to investigate the value of strain rate ratio (SR) of elastography for benign and malignant breast masses categorized by BI-RADS for Ultrasonography (BI-RADS-US) category 3-4. Ultrasonographic data of 1099 breast masses that underwent both elastography and pathological examination were retrospectively analyzed by non-parametric test, consistency analysis and receiver operating characteristic (ROC) curve sequentially. ROC curve was used to evaluate the diagnostic effect of SR method in different BI-RADS categories of breast lesions. The maximum Youden index obtained from the ROC curve was 0.845. At a cut-off value of 3.57, the SR values' diagnostic sensitivity, specificity, accuracy, positive predictive value and negative predictive value for distinguishing benign and malignant breast lesions were 85.7 %, 98.8 %, 93.6 %, 97.9 % and 91.45 %, respectively. Consistency analysis showed that the consistency between SR value and pathological diagnosis was 93.6 % (κ = 0.864). In addition, the SR values between benign and malignant lesions of BI-RADS 3 and 4 were statistically different (P < 0.001). ROC analysis indicated that the diagnostic area under curve (AUC) for SR value in BI-RADS 3, 4a, 4b and 4c lesions were 0.985, 0.866, 0.793 and 0.916, respectively. In addition, the study observed 58 cases of missed diagnosis and 20 cases of misdiagnosis in evaluating benign and malignant breast lesions. The elastic SR ratio method has a good diagnostic value for the evaluation of breast masses, particularly for lesions with a score of 3-4. The elastic SR ratio method is instrumental in enhancing the accuracy of diagnosis.

Abstract GS01-06: Advancing Evidence of the Associations Between Specific Benign Breast Diagnoses and Future Breast Cancer RIsk

Abstract Background: Benign breast disease (BBD) is a common breast biopsy finding encompassing a diverse spectrum of diagnoses. Breast cancer risk associated with many specific BBD diagnoses and their joint associations with breast density have not been extensively studied. We estimate the future risk of invasive breast cancer associated with specific BBD diagnoses typically categorized into the broad category of proliferative lesions without atypia (PWoA) and evaluate whether associations differ by breast density or by presence of calcifications. Methods: We included 1,313,943 women ages 40-79 years who underwent 5,282,063 mammograms in the Breast Cancer Surveillance Consortium from 1996 to 2019 with no prior history of breast cancer. We fit a Cox proportional hazards model to estimate hazard ratios associated with each combination of PWoA BBD diagnoses, whether calcifications were present on biopsy, and breast density. We used classification trees to group combinations of PWoA diagnoses, breast density, age groups, and presence of calcifications with similar magnitudes of associations with 5-year risk of invasive breast cancer. Results: Compared with women with a PWoA BBD diagnosis without calcifications, presence of calcifications significantly elevated risk for breast cancer (HR=1.19, 95% CI = 1.02 to 1.41, P=0.03). Considering specific PWoA diagnoses, classification trees identified groups of women with higher and lower invasive cancer risk than expected risk for all PWoA combined. Among women ages 60-79 years with heterogeneously or extremely dense breasts, the 2.6% of women with any PWoA diagnosis, except papillomatosis, were at high invasive breast cancer risk (5-year risk &amp;gt;2.5%) (Table 1). Among women ages 40-59 years with dense breasts, presence of calcifications with papillomas, usual ductal hyperplasia, columnar cell hyperplasia, or PWoA NOS diagnoses put these women at high risk (5-year risk &amp;gt;2.5%) compared to women with the same BBD diagnoses without calcifications who were at intermediate or average risk (1-2.49%) (Table 1). Among women ages 60-79 years with non-dense breasts, the 1.6% of women with papilloma, usual ductal hyperplasia, radial scar, or PWoA NOS diagnoses were at high breast cancer risk (Table 2). Among women ages 40-59 years with non-dense breasts, presence of calcifications with usual ductal hyperplasia or PwoA NOS diagnoses put these women at intermediate risk (1.67-2.49%) compared to women with these BBD diagnoses without calcifications who were at average risk (1-1.66%) (Table 2). Women 60-79 years with fatty breasts were at intermediate 5-year risk if they had papillomas, usual ductal hyperplasia, or PWoA NOS diagnoses. Women 40-59 years with fatty breasts were at low risk of breast cancer regardless of PWoA diagnosis (Table 2). Conclusion: Specific BBD diagnoses and the presence of calcifications can change a woman’s predicted 5-year breast cancer risk compared to broad BBD categories alone. This information could be incorporated into risk prediction models to improve model accuracy. Citation Format: Olivia Sattayapiwat, Karla Kerlikowske, Donald Weaver, Alexander Borowsky, Theresa Keegan, Brian Sprague, Diana Miglioretti. Advancing Evidence of the Associations Between Specific Benign Breast Diagnoses and Future Breast Cancer RIsk [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr GS01-06.

Thoracoscopic segmentectomy for trans-fissure ground-glass opacity.

The trans-fissure ground-glass opacity (GGO) is a special category of lesions, with a diameter always exceeding 2 cm. It is located on a fused fissure, 'seizing' 2 neighbouring lobes simultaneously. The segmentectomy for the trans-fissure GGO is never reported. Between August 2016 and December 2022, patients operated with a trans-fissure GGO were included. The patients' backgrounds and surgical data were summarized. All procedures were performed with the help of preoperative three-dimensional computed tomography bronchography and angiography. A total of 84 patients were included. The selection criteria included a consolidation tumour ratio <50% and a lesion size >2 and ≤3 cm. Thirty-six patients were operated with lobectomy + wedge (the traditional method group) and 48 patients were operated with anatomical segmentectomy + function-preserving sublobectomy (the new method group). The median operative time was 87 min in the traditional group and 98 min in the new method group, and the median blood loss was 60 ml in the traditional group and 70 ml in the new method group. The median duration of hospital stays was 4 days in the traditional group and 2 days in the new method group. In the traditional method group, there was 1 case of postoperative air leakage and 5 cases of haemoptysis. In the new method group, 2 cases of postoperative air leakage were identified. The median size of the tumour in the resected segment was 2.6 cm in the traditional group and 2.5 cm in the new method group. The median margin was 2.5 cm in the traditional group and 3.3 cm in the new method group. The trans-fissure GGO could be safely resected en bloc by segmentectomy with a well-designed surgical procedure and appropriate preoperative planning.

Explainable DCNN Decision Framework for Breast Lesion Classification from Ultrasound Images Based on Cancer Characteristics.

In recent years, deep convolutional neural networks (DCNNs) have shown promising performance in medical image analysis, including breast lesion classification in 2D ultrasound (US) images. Despite the outstanding performance of DCNN solutions, explaining their decisions remains an open investigation. Yet, the explainability of DCNN models has become essential for healthcare systems to accept and trust the models. This paper presents a novel framework for explaining DCNN classification decisions of lesions in ultrasound images using the saliency maps linking the DCNN decisions to known cancer characteristics in the medical domain. The proposed framework consists of three main phases. First, DCNN models for classification in ultrasound images are built. Next, selected methods for visualization are applied to obtain saliency maps on the input images of the DCNN models. In the final phase, the visualization outputs and domain-known cancer characteristics are mapped. The paper then demonstrates the use of the framework for breast lesion classification from ultrasound images. We first follow the transfer learning approach and build two DCNN models. We then analyze the visualization outputs of the trained DCNN models using the EGrad-CAM and Ablation-CAM methods. We map the DCNN model decisions of benign and malignant lesions through the visualization outputs to the characteristics such as echogenicity, calcification, shape, and margin. A retrospective dataset of 1298 US images collected from different hospitals is used to evaluate the effectiveness of the framework. The test results show that these characteristics contribute differently to the benign and malignant lesions' decisions. Our study provides the foundation for other researchers to explain the DCNN classification decisions of other cancer types.

Abstract PO1-07-03: An Accurate Breast Cancer Diagnosis Method Using a Multi-Feature Fusion Neural Network on Contrast-Enhanced Spectral Mammography

Abstract Objectives: Our study aims to develop an accurate and comprehensive deep neural network capable of classifying CESM images to aid in the early detection and diagnosis of breast cancer in clinical settings. Methods: We enrolled diagnostic CESM examinations conducted between January 1, 2019 and January 17, 2021. We developed and tested the performance of a multi-feature fusion network for breast lesion classification by combining low-energy (LE) and dual-energy subtracted (DES) images, dual-view, and bilateral information using a large and diverse CESM dataset. We have evaluated the ability of the proposed network to generalize on external datasets and the results were reported using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity. Results: In the study period 1973CESMs were included. Mean age was 53 years ± 12 (standard deviation). In the internal test dataset, the model with CESM (LE +DES) inputs combined not only bilateral information (left and right breasts) but also dualview information (CC and MLO), achieved best diagnosis performance with AUC of 0.96 [95% confidence interval (CI): 0.95–0.97], accuracy of 0.9 [95% confidence interval (CI): 0.88–0.92], sensitivity of 0.84 [95% confidence interval (CI): 0.80–0.88] and specificity of 0.93 [95% confidence interval (CI): 0.91–0.95], while the model with LE inputs only got an AUC of 0.94 [95% (CI): 0.93–0.95], accuracy of 0.88 [95% confidence interval (CI): 0.86–0.90], sensitivity of 0.79 [95% confidence interval (CI): 0.74–0.83] and specificity of 0.93 [95% confidence interval (CI): 0.91–0.95] (external dataset: an AUC of 0.90 [95% (CI): 0.85–0.94], accuracy of 0.84 [95% confidence interval (CI): 0.78–0.89], sensitivity of 0.77 [95% confidence interval (CI): 0.61–0.88] and specificity of 0.87 [95% confidence interval (CI):0.80–0.92]). Conclusion: CESM is a promising technique in breast cancer diagnosis due to its high feasibility and potential. The results demonstrate that left-right breast fusion and dual-view fusion are helpful for accurate diagnosis on CESM images. Specifically, dual-energy subtracted (DES) images generated by CESM improve diagnostic accuracy when compared to low-energy (LE) images alone. Keywords: Contrast-enhanced spectral mammography; Deep neural network; Breast cancer diagnosis; Multi-feature fusion Citation Format: Yipeng Song, Wei Jiang. An Accurate Breast Cancer Diagnosis Method Using a Multi-Feature Fusion Neural Network on Contrast-Enhanced Spectral Mammography [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-07-03.

TransEBUS: The interpretation of endobronchial ultrasound image using hybrid transformer for differentiating malignant and benign mediastinal lesions

The purpose of this study is to establish a deep learning automatic assistance diagnosis system for benign and malignant classification of mediastinal lesions in endobronchial ultrasound (EBUS) images. EBUS images are in the form of video and contain multiple imaging modes. Different imaging modes and different frames can reflect the different characteristics of lesions. Compared with previous studies, the proposed model can efficiently extract and integrate the spatiotemporal relationships between different modes and does not require manual selection of representative frames. In recent years, Vision Transformer has received much attention in the field of computer vision. Combined with convolutional neural networks, hybrid transformers can also perform well on small datasets. This study designed a novel deep learning architecture based on hybrid transformer called TransEBUS. By adding learnable parameters in the temporal dimension, TransEBUS was able to extract spatiotemporal features from insufficient data. In addition, we designed a two-stream module to integrate information from three different imaging modes of EBUS. Furthermore, we applied contrastive learning when training TransEBUS, enabling it to learn discriminative representation of benign and malignant mediastinal lesions. The results show that TransEBUS achieved a diagnostic accuracy of 82% and an area under the curve of 0.8812 in the test dataset, outperforming other methods. It also shows that several models can improve performance by incorporating two-stream module. Our proposed system has shown its potential to help physicians distinguishing benign and malignant mediastinal lesions, thereby ensuring the accuracy of EBUS examination.

Artificial intelligence-based classification of breast lesion from contrast enhanced mammography: a multicenter study.

The authors aimed to establish an artificial intelligence (AI)-based method for preoperative diagnosis of breast lesions from contrast enhanced mammography (CEM) and to explore its biological mechanism. This retrospective study includes 1430 eligible patients who underwent CEM examination from June 2017 to July 2022 and were divided into a construction set (n=1101), an internal test set (n=196), and a pooled external test set (n=133). The AI model adopted RefineNet as a backbone network, and an attention sub-network, named convolutional block attention module (CBAM), was built upon the backbone for adaptive feature refinement. An XGBoost classifier was used to integrate the refined deep learning features with clinical characteristics to differentiate benign and malignant breast lesions. The authors further retrained the AI model to distinguish in situ and invasive carcinoma among breast cancer candidates. RNA-sequencing data from 12 patients were used to explore the underlying biological basis of the AI prediction. The AI model achieved an area under the curve of 0.932 in diagnosing benign and malignant breast lesions in the pooled external test set, better than the best-performing deep learning model, radiomics model, and radiologists. Moreover, the AI model has also achieved satisfactory results (an area under the curve from 0.788 to 0.824) for the diagnosis of in situ and invasive carcinoma in the test sets. Further, the biological basis exploration revealed that the high-risk group was associated with the pathways such as extracellular matrix organization. The AI model based on CEM and clinical characteristics had good predictive performance in the diagnosis of breast lesions.

Molecular Classification of Vulvar Squamous Cell Carcinoma and Precursor Lesions Using Immunohistochemistry

Abstract Introduction: Vulvar malignancies are very rare. Two pathogenic pathways have been described in the development of vulvar squamous cell carcinoma.One is related to HPV infection and the other is related to TP 53 mutation. HPV infection related squamous cell carcinomas have a good prognosis and squamous cell carcinomas related to TP 53 mutation have the worst prognosis. Recently a third pathway which is neither HPV dependent nor p53 mutant has been identified. This pathway is found to have intermediate prognosis with better clinical outcome. Aims: The aim of this study was to classify vulvar squamous cell carcinoma and its precursor lesions molecularly using two immunohistochemical stains p16 and p53 in a cohort of South Indian population. Subjects and Methods: A retrospective observational study was conducted between January 2011 and December 2020 in the department of histopathology at a tertiary care hospital in Hyderabad, India. A total of 22 cases including all the vulval biopsies with suspected squamous cell carcinoma or precursor lesions and excision specimens operated for squamous cell carcinoma were included in this study. Paraffin blocks of representative cases were stained with hematoxylin and eosin stain, p16 and p53 immunohistochemistry (IHC) stains. Results: Out of 22 cases, two cases (9.0%) were precursor lesions of the vulva and 20 cases (90.9%) were invasive squamous cell carcinoma. Eleven (50%) of these cases were p53 mutant (p53+/p16−), 10 (45.4%) of these cases were HPV positive (p16+/p53−), and one of these cases (4.5%) was neither HPV positive nor p53 mutant (p16−/p53−). Conclusions: Classification of vulvar squamous cell lesions molecularly using simple immunohistochemical stains in routine diagnostic workup will aid in risk stratification of patients and help in their management and prognostication.

Evaluation of pap smears with cyto-histopathological correlation–our experience at a tertiary care hospital in Telangana

Background: The World Health Organization states that cervical carcinoma is the fourth most common cancer among women worldwide. Pap smear is the simplest and easiest test to detect this prevalent cancer at its nascent stages. Aims and Objectives: This study aims to evaluate Pap smears using the conventional method (Bethesda 2014) and correlate the abnormal smears with cervical biopsy. Materials and Methods: The study is a combined retrospective and prospective study, from January 2021 to August 2023. Pap smears received were evaluated according to Bethesda 2014. Those cases categorized as Epithelial Cell Abnormality (ECA), were correlated with corresponding histopathology on cervical biopsy. Results: Out of the 615 pap smears in this study, 525 cases (91%) were categorized as Negative for Intraepithelial Lesion or Malignancy, 53 cases (9%) as Epithelial Cell Abnormality, and 37 cases (6%) as unsatisfactory. In the Negative for Intraepithelial Lesion or Malignancy category, 353 cases (67.2%) were reactive cellular changes including inflammation, four cases (0.7%) were squamous metaplasia, and 24 cases (4.6%) were atrophic smears. In the ECA category–21 cases (39.6%) were Atypical squamous cells of undetermined significance, two cases (3.8%) were Low-grade squamous intraepithelial lesion, five cases (9.4%) were Atypical squamous cells-cannot exclude HSIL(ASC-H), 14 cases (26.4%) were High grade squamous intraepithelial lesion (HSIL) and 3 cases (5.6%) were squamous cell carcinoma (SCC) cervix, seven cases were of ECA-glandular type (13.2%). On histopathological correlation, the maximum correlation was found for HSIL and SCC cases, 71% and 100%, respectively. Conclusion: Pap smear is a crucial screening tool to diagnose cervical carcinoma at its nascent stage. This study showed a positive correlation for high-grade lesions on Pap smears.

A relationship-aware mutual learning method for lightweight skin lesion classification

A relationship-aware mutual learning method for lightweight skin lesion classification