Automated assessment of breast cancer margin in optical coherence tomography images via pretrained convolutional neural network

Comparing the diagnostic efficacy of optical coherence tomography and frozen section for margin assessment in breast-conserving surgery: a meta-analysis

AIMS

This meta-analysis assessed the relative diagnostic accuracy of optical coherence tomography (OCT) versus frozen section (FS) in evaluating surgical margins during breast-conserving procedures.

METHODS

PubMed and Embase were searched for relevant studies published up to October 2023. The inclusion criteria encompassed studies evaluating the diagnostic accuracy of OCT or FS in patients undergoing breast-conserving surgery. Sensitivity and specificity were analysed using the DerSimonian and Laird method and subsequently transformed through the Freeman-Tukey double inverse sine method.

RESULTS

The meta-analysis encompassed 36 articles, comprising 16 studies on OCT and 20 on FS, involving 10 289 specimens from 8058 patients. The overall sensitivity of OCT was 0.93 (95% CI: 0.90 to 0.96), surpassing that of FS, which was 0.82 (95% CI: 0.71 to 0.92), indicating a significantly higher sensitivity for OCT (p=0.04). Conversely, the overall specificity of OCT was 0.89 (95% CI: 0.83 to 0.94), while FS exhibited a higher specificity at 0.97 (95% CI: 0.95 to 0.99), suggesting a superior specificity for FS (p＜0.01).

CONCLUSIONS

Our meta-analysis reveals that OCT offers superior sensitivity but inferior specificity compared with FS in assessing surgical margins in breast-conserving surgery patients. Further larger well-designed prospective studies are needed, especially those employing a head-to-head comparison design.

PROSPERO REGISTRATION NUMBER

CRD42023483751.

Optical Imaging in Human Lymph Node Specimens for Detecting Breast Cancer Metastases: A Review

Assessment of regional lymph node status in breast cancer is of important staging and prognostic value. Even though formal histological examination is the currently accepted standard of care, optical imaging techniques have shown promising results in disease diagnosis. In the present article, we review six spectroscopic techniques and focus on their use as alternative tools for breast cancer lymph node assessment. Elastic scattering spectroscopy (ESS) seems to offer a simple, cost-effective, and reproducible method for intraoperative diagnosis of breast cancer lymph node metastasis. Optical coherence tomography (OCT) provides high-resolution tissue scanning, along with a short data acquisition time. However, it is relatively costly and experimentally complex. Raman spectroscopy proves to be a highly accurate method for the identification of malignant axillary lymph nodes, and it has been further validated in the setting of head and neck cancers. Still, it remains time-consuming. Near-infrared fluorescence imaging (NIRF) and diffuse reflectance spectroscopy (DFS) are related to significant advantages, such as deep tissue penetration and efficiency. Fourier-transform infrared spectroscopy (FTIR) is a promising method but has significant drawbacks. Nonetheless, only anecdotal reports exist on their clinical use for cancerous lymph node detection. Our results indicate that optical imaging methods can create informative and rapid tools to effectively guide surgical decision-making.

Rapid On-Site AI-Assisted Grading for Lung Surgery Based on Optical Coherence Tomography

The determination of resection extent traditionally relies on the microscopic invasiveness of frozen sections (FSs) and is crucial for surgery of early lung cancer with preoperatively unknown histology. While previous research has shown the value of optical coherence tomography (OCT) for instant lung cancer diagnosis, tumor grading through OCT remains challenging. Therefore, this study proposes an interactive human-machine interface (HMI) that integrates a mobile OCT system, deep learning algorithms, and attention mechanisms. The system is designed to mark the lesion's location on the image smartly and perform tumor grading in real time, potentially facilitating clinical decision making. Twelve patients with a preoperatively unknown tumor but a final diagnosis of adenocarcinoma underwent thoracoscopic resection, and the artificial intelligence (AI)-designed system mentioned above was used to measure fresh specimens. Results were compared to FSs benchmarked on permanent pathologic reports. Current results show better differentiating power among minimally invasive adenocarcinoma (MIA), invasive adenocarcinoma (IA), and normal tissue, with an overall accuracy of 84.9%, compared to 20% for FSs. Additionally, the sensitivity and specificity, the sensitivity and specificity were 89% and 82.7% for MIA and 94% and 80.6% for IA, respectively. The results suggest that this AI system can potentially produce rapid and efficient diagnoses and ultimately improve patient outcomes.

Diagnostic accuracy of optical coherence tomography for margin assessment in breast-conserving surgery: A systematic review and meta-analysis

BACKGROUND

Breast cancer is the most common malignant tumor among women, and its incidence is increasing annually. At present, the results of the study on whether optical coherence tomography (OCT) can be used as an intraoperative margin assessment method for breast-conserving surgery (BCS) are inconsistent. We herein conducted this systematic review and meta-analysis to assess the diagnostic value of OCT in BCS.

METHODS

PubMed, Web of Science, Cochrane Library, and Embase were used to search relevant studies published up to September 15, 2022. We used Review Manager 5.4, Meta-Disc 1.4, and STATA 16.0 for statistical analysis.

RESULTS

The results displayed 18 studies with 782 patients included according to the inclusion and exclusion criteria. Meta-analysis showed the pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR) and the area under the curve (AUC) of OCT in the margin assessment of BCS were 0.91 (95% CI 0.88-0.93), 0.88 (95% CI 0.83-0.92), 7.53 (95% CI 5.19-10.93), 0.11(95% CI 0.08-0.14), 70.37 (95% CI 39.78-124.47), and 0.94 (95% CI 0.92-0.96), respectively.

CONCLUSIONS

OCT is a promising technique in intraoperative margin assessment of breast cancer patients.

Diseased thyroid tissue classification in OCT images using deep learning: Towards surgical decision support

Intraoperative guidance tools for thyroid surgery based on optical coherence tomography (OCT) could aid distinguish between normal and diseased tissue. However, OCT images are difficult to interpret, thus, real-time automatic analysis could support the clinical decision-making. In this study, several deep learning models were investigated for thyroid disease classification on 2D and 3D OCT data obtained from ex vivo specimens of 22 patients undergoing surgery and diagnosed with several thyroid pathologies. Additionally, two open-access datasets were used to evaluate the custom models. On the thyroid dataset, the best performance was achieved by the 3D vision transformer model with a Matthew's correlation coefficient (MCC) of 0.79 (accuracy = 0.90) for the normal-versus-abnormal classification. On the open-access datasets, the custom models achieved the best performance (MCC > 0.88, accuracy > 0.96). Results obtained for the normal-versus-abnormal classification suggest OCT, complemented with deep learning-based analysis, as a tool for real-time automatic diseased tissue identification in thyroid surgery.

Inflation of test accuracy due to data leakage in deep learning-based classification of OCT images

In the application of deep learning on optical coherence tomography (OCT) data, it is common to train classification networks using 2D images originating from volumetric data. Given the micrometer resolution of OCT systems, consecutive images are often very similar in both visible structures and noise. Thus, an inappropriate data split can result in overlap between the training and testing sets, with a large portion of the literature overlooking this aspect. In this study, the effect of improper dataset splitting on model evaluation is demonstrated for three classification tasks using three OCT open-access datasets extensively used, Kermany's and Srinivasan's ophthalmology datasets, and AIIMS breast tissue dataset. Results show that the classification performance is inflated by 0.07 up to 0.43 in terms of Matthews Correlation Coefficient (accuracy: 5% to 30%) for models tested on datasets with improper splitting, highlighting the considerable effect of dataset handling on model evaluation. This study intends to raise awareness on the importance of dataset splitting given the increased research interest in implementing deep learning on OCT data.

Binary dose level classification of tumour microvascular response to radiotherapy using artificial intelligence analysis of optical coherence tomography images

The dominant consequence of irradiating biological systems is cellular damage, yet microvascular damage begins to assume an increasingly important role as the radiation dose levels increase. This is currently becoming more relevant in radiation medicine with its pivot towards higher-dose-per-fraction/fewer fractions treatment paradigm (e.g., stereotactic body radiotherapy (SBRT)). We have thus developed a 3D preclinical imaging platform based on speckle-variance optical coherence tomography (svOCT) for longitudinal monitoring of tumour microvascular radiation responses in vivo. Here we present an artificial intelligence (AI) approach to analyze the resultant microvascular data. In this initial study, we show that AI can successfully classify SBRT-relevant clinical radiation dose levels at multiple timepoints (t = 2–4 weeks) following irradiation (10 Gy and 30 Gy cohorts) based on induced changes in the detected microvascular networks. Practicality of the obtained results, challenges associated with modest number of animals, their successful mitigation via augmented data approaches, and advantages of using 3D deep learning methodologies, are discussed. Extension of this encouraging initial study to longitudinal AI-based time-series analysis for treatment outcome predictions at finer dose level gradations is envisioned.

Research progress on the application of optical coherence tomography in the field of oncology

Optical coherence tomography (OCT) is a non-invasive imaging technique which has become the “gold standard” for diagnosis in the field of ophthalmology. However, in contrast to the eye, nontransparent tissues exhibit a high degree of optical scattering and absorption, resulting in a limited OCT imaging depth. And the progress made in the past decade in OCT technology have made it possible to image nontransparent tissues with high spatial resolution at large (up to 2mm) imaging depth. On the one hand, OCT can be used in a rapid, noninvasive way to detect diseased tissues, organs, blood vessels or glands. On the other hand, it can also identify the optical characteristics of suspicious parts in the early stage of the disease, which is of great significance for the early diagnosis of tumor diseases. Furthermore, OCT imaging has been explored for imaging tumor cells and their dynamics, and for the monitoring of tumor responses to treatments. This review summarizes the recent advances in the OCT area, which application in oncological diagnosis and treatment in different types: (1) superficial tumors:OCT could detect microscopic information on the skin’s surface at high resolution and has been demonstrated to help diagnose common skin cancers; (2) gastrointestinal tumors: OCT can be integrated into small probes and catheters to image the structure of the stomach wall, enabling the diagnosis and differentiation of gastrointestinal tumors and inflammation; (3) deep tumors: with the rapid development of OCT imaging technology, it has shown great potential in the diagnosis of deep tumors such in brain tumors, breast cancer, bladder cancer, and lung cancer.

Multi-class classification of breast tissue using optical coherence tomography and attenuation imaging combined via deep learning

We demonstrate a convolutional neural network (CNN) for multi-class breast tissue classification as adipose tissue, benign dense tissue, or malignant tissue, using multi-channel optical coherence tomography (OCT) and attenuation images, and a novel Matthews correlation coefficient (MCC)-based loss function that correlates more strongly with performance metrics than the commonly used cross-entropy loss. We hypothesized that using multi-channel images would increase tumor detection performance compared to using OCT alone. 5,804 images from 29 patients were used to fine-tune a pre-trained ResNet-18 network. Adding attenuation images to OCT images yields statistically significant improvements in several performance metrics, including benign dense tissue sensitivity (68.0% versus 59.6%), malignant tissue positive predictive value (PPV) (79.4% versus 75.5%), and total accuracy (85.4% versus 83.3%), indicating that the additional contrast from attenuation imaging is most beneficial for distinguishing between benign dense tissue and malignant tissue.

Intraoperative Margin Trials in Breast Cancer

Purpose of Review

Obtaining negative margins in breast conservation surgery continues to be a challenge. Re-excisions are difficult for patients and expensive for the health systems. This paper reviews the literature on current strategies and intraoperative clinical trials to reduce positive margin rates.

Recent Findings

The best available data demonstrate that intraoperative imaging with ultrasound, intraoperative pathologic assessment such as frozen section, and cavity margins have been the most successful intraoperative strategies to reduce positive margins. Emerging technologies such as optical coherence tomography and fluorescent imaging need further study but may be important adjuncts.

Summary

There are several proven strategies to reduce positive margin rates to < 10%. Surgeons should utilize best available resources within their institutions to produce the best outcomes for their patients.

Classifying breast cancer in ultrahigh-resolution optical coherence tomography images using convolutional neural networks

Optical coherence tomography (OCT) is being investigated in breast cancer diagnostics as a real-time histology evaluation tool. We present a customized deep convolutional neural network (CNN) for classification of breast tissues in OCT B-scans. Images of human breast samples from mastectomies and breast reductions were acquired using a custom ultrahigh-resolution OCT system with 2.72 µm axial resolution and 5.52 µm lateral resolution. The network achieved 96.7% accuracy, 92% sensitivity, and 99.7% specificity on a dataset of 23 patients. The usage of deep learning will be important for the practical integration of OCT into clinical practice.

Transfer learning for medical image classification: a literature review

BACKGROUND

Transfer learning (TL) with convolutional neural networks aims to improve performances on a new task by leveraging the knowledge of similar tasks learned in advance. It has made a major contribution to medical image analysis as it overcomes the data scarcity problem as well as it saves time and hardware resources. However, transfer learning has been arbitrarily configured in the majority of studies. This review paper attempts to provide guidance for selecting a model and TL approaches for the medical image classification task.

METHODS

425 peer-reviewed articles were retrieved from two databases, PubMed and Web of Science, published in English, up until December 31, 2020. Articles were assessed by two independent reviewers, with the aid of a third reviewer in the case of discrepancies. We followed the PRISMA guidelines for the paper selection and 121 studies were regarded as eligible for the scope of this review. We investigated articles focused on selecting backbone models and TL approaches including feature extractor, feature extractor hybrid, fine-tuning and fine-tuning from scratch.

RESULTS

The majority of studies (n = 57) empirically evaluated multiple models followed by deep models (n = 33) and shallow (n = 24) models. Inception, one of the deep models, was the most employed in literature (n = 26). With respect to the TL, the majority of studies (n = 46) empirically benchmarked multiple approaches to identify the optimal configuration. The rest of the studies applied only a single approach for which feature extractor (n = 38) and fine-tuning from scratch (n = 27) were the two most favored approaches. Only a few studies applied feature extractor hybrid (n = 7) and fine-tuning (n = 3) with pretrained models.

CONCLUSION

The investigated studies demonstrated the efficacy of transfer learning despite the data scarcity. We encourage data scientists and practitioners to use deep models (e.g. ResNet or Inception) as feature extractors, which can save computational costs and time without degrading the predictive power.

OCT-Guided Surgery for Gliomas: Current Concept and Future Perspectives

Optical coherence tomography (OCT) has been recently suggested as a promising method to obtain in vivo and real-time high-resolution images of tissue structure in brain tumor surgery. This review focuses on the basics of OCT imaging, types of OCT images and currently suggested OCT scanner devices and the results of their application in neurosurgery. OCT can assist in achieving intraoperative precision identification of tumor infiltration within surrounding brain parenchyma by using qualitative or quantitative OCT image analysis of scanned tissue. OCT is able to identify tumorous tissue and blood vessels detection during stereotactic biopsy procedures. The combination of OCT with traditional imaging such as MRI, ultrasound and 5-ALA fluorescence has the potential to increase the safety and accuracy of the resection. OCT can improve the extent of resection by offering the direct visualization of tumor with cellular resolution when using microscopic OCT contact probes. The theranostic implementation of OCT as a part of intelligent optical diagnosis and automated lesion localization and ablation could achieve high precision, automation and intelligence in brain tumor surgery. We present this review for the increase of knowledge and formation of critical opinion in the field of OCT implementation in brain tumor surgery.

Development of intraoperative assessment of margins in breast conserving surgery: a narrative review

OBJECTIVE

We intend to provide an informative and up-to-date summary on the topic of intraoperative assessment of margins in breast conserving surgery (BCS). Conventional methods as well as cutting-edge technologies are analyzed for their advantages and limitations in the hope that clinicians can turn to this for reference. This review can also offer guidance for technicians in the future design of intraoperative margin assessment tools.

BACKGROUND

Achieving negative margins during BCS is one of the vital factors for preventing local recurrence. Conducting intraoperative margin assessment can ensure negative margins to a large extent and possibly relieve patients of the anguish of re-interventions. In recent years, innovative methods for margin assessment during BCS are advancing rapidly. And there is a lack of summary regarding the development of intraoperative margin assessment in BCS.

METHODS

A PubMed search with keywords "intraoperative margin assessment" and "breast conserving surgery" was conducted. Relevant publications were screened manually for its title, abstract and even full text to determine its true relevance. Publications on neo-adjuvant therapy and intraoperative radiotherapy were excluded. References from the searched articles and other supplementary articles were also looked into.

CONCLUSIONS

Conventional methods for margin assessment yields stable outcome but its use is limited because of the demand on pathology staff and the trade-off between time and precision. Conventional imaging techniques pass the workload to radiologists at the cost of a significantly low duration of time. Involving artificial intelligence for image-based assessment is a further improvement. However, conventional imaging is inherently flawed in that occult lesions can't show on the image and the showing ones are ambiguous and open to interpretation. Unconventional techniques which base their judgment on cellular composition are more reassuring. Nonetheless, unconventional techniques should be subjected to clinical trials before putting into practice. And studies regarding comparison between conventional methods and unconventional methods are also needed to evaluate their relative efficacy.

Optical coherence tomography for identification of malignant pulmonary nodules based on random forest machine learning algorithm

OBJECTIVE

To explore the feasibility of using random forest (RF) machine learning algorithm in assessing normal and malignant peripheral pulmonary nodules based on in vivo endobronchial optical coherence tomography (EB-OCT).

METHODS

A total of 31 patients with pulmonary nodules were admitted to Department of Respiratory Medicine, Zhongda Hospital, Southeast University, and underwent chest CT, EB-OCT and biopsy. Attenuation coefficient and up to 56 different image features were extracted from A-line and B-scan of 1703 EB-OCT images. Attenuation coefficient and 29 image features with significant p-values were used to analyze the differences between normal and malignant samples. A RF classifier was trained using 70% images as training set, while 30% images were included in the testing set. The accuracy of the automated classification was validated by clinically proven pathological results.

RESULTS

Attenuation coefficient and 29 image features were found to present different properties with significant p-values between normal and malignant EB-OCT images. The RF algorithm successfully classified the malignant pulmonary nodules with sensitivity, specificity, and accuracy of 90.41%, 77.87% and 83.51% respectively.

CONCLUSION

It is clinically practical to distinguish the nature of pulmonary nodules by integrating EB-OCT imaging with automated machine learning algorithm. Diagnosis of malignant pulmonary nodules by analyzing quantitative features from EB-OCT images could be a potentially powerful way for early detection of lung cancer.

Classification of cell morphology with quantitative phase microscopy and machine learning

We describe and compare two machine learning approaches for cell classification based on label-free quantitative phase imaging with transport of intensity equation methods. In one approach, we design a multilevel integrated machine learning classifier including various individual models such as artificial neural network, extreme learning machine and generalized logistic regression. In another approach, we apply a pretrained convolutional neural network using transfer learning for the classification. As a validation, we show the performances of both approaches on classification between macrophages cultured in normal gravity and microgravity with quantitative phase imaging. The multilevel integrated classifier achieves average accuracy 93.1%, which is comparable to the average accuracy 93.5% obtained by convolutional neural network. The presented quantitative phase imaging system with two classification approaches could be helpful to biomedical scientists for easy and accurate cell analysis.

Deep learning a boon for biophotonics?

This review covers original articles using deep learning in the biophotonic field published in the last years. In these years deep learning, which is a subset of machine learning mostly based on artificial neural network geometries, was applied to a number of biophotonic tasks and has achieved state-of-the-art performances. Therefore, deep learning in the biophotonic field is rapidly growing and it will be utilized in the next years to obtain real-time biophotonic decision-making systems and to analyze biophotonic data in general. In this contribution, we discuss the possibilities of deep learning in the biophotonic field including image classification, segmentation, registration, pseudostaining and resolution enhancement. Additionally, we discuss the potential use of deep learning for spectroscopic data including spectral data preprocessing and spectral classification. We conclude this review by addressing the potential applications and challenges of using deep learning for biophotonic data.

Insights into Hypoxia: Non-invasive Assessment through Imaging Modalities and Its Application in Breast Cancer

Oxygen is crucial to maintain the homeostasis in aerobic cells. Hypoxia is a condition in which cells are deprived of the oxygen supply necessary for their optimum performance. Whereas oxygen deprivation may occur in normal physiological processes, hypoxia is frequently associated with pathological conditions. It has been identified as a stressor in the tumor microenvironment, acting as a key mediator of cancer development. Numerous pathways are activated in hypoxic cells that affect cell signaling and gene regulation to promote the survival of these cells by stimulating angiogenesis, switching cellular metabolism, slowing their growth rate, and preventing apoptosis. The induction of dysregulated metabolism in cancer cells by hypoxia results in aggressive tumor phenotypes that are characterized by rapid progression, treatment resistance, and poor prognosis. A non-invasive assessment of hypoxia-induced metabolic and architectural changes in tumors is advisable to fully improve breast cancer (BC) patient management, by potentially reducing the need for invasive biopsy procedures and evaluating tumor response to treatment. This review provides a comprehensive overview of the molecular changes in breast tumors secondary to hypoxia and the non-invasive imaging alternatives to evaluate oxygen deprivation, with an emphasis on their application in BC and the advantages and limitations of the currently available techniques.