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

Scientific Data, Journal Year: 2022, Volume and Issue: 9(1)

Published: Sept. 22, 2022

Abstract 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 micrometer resolution OCT systems, consecutive are often very similar in both visible structures and noise. Thus, an inappropriate data split can result overlap between training testing sets, with a large portion literature overlooking this aspect. study, effect improper dataset splitting model evaluation demonstrated for three tasks open-access datasets extensively used, Kermany’s Srinivasan’s ophthalmology datasets, AIIMS breast tissue dataset. Results show that performance inflated by 0.07 up 0.43 terms Matthews Correlation Coefficient (accuracy: 5% 30%) models tested splitting, highlighting considerable handling evaluation. This study intends raise awareness importance given increased research interest implementing

Language: Английский

---

Computer-aided diagnosis of retinopathy based on vision transformer

Journal of Innovative Optical Health Sciences, Journal Year: 2021, Volume and Issue: 15(02)

Published: Nov. 29, 2021

Age-related Macular Degeneration (AMD) and Diabetic Edema (DME) are two common retinal diseases for elder people that may ultimately cause irreversible blindness. Timely accurate diagnosis is essential the treatment of these diseases. In recent years, computer-aided (CAD) has been deeply investigated effectively used rapid early diagnosis. this paper, we proposed a method CAD using vision transformer to analyze optical coherence tomography (OCT) images automatically discriminate AMD, DME, normal eyes. A classification accuracy 99.69% was achieved. After model pruning, recognition time reached 0.010 s did not drop. Compared with Convolutional Neural Network (CNN) image models (VGG16, Resnet50, Densenet121, EfficientNet), after pruning exhibited better ability. Results show an improved alternative diagnose more accurately.

Language: Английский

---

Extended continuous similarity indices: theory and application for QSAR descriptor selection

Journal of Computer-Aided Molecular Design, Journal Year: 2022, Volume and Issue: 36(3), P. 157 - 173

Published: March 1, 2022

Language: Английский

---

Towards targeted colorectal cancer biopsy based on tissue morphology assessment by compression optical coherence elastography

Frontiers in Oncology, Journal Year: 2023, Volume and Issue: 13

Published: March 27, 2023

Identifying the precise topography of cancer for targeted biopsy in colonoscopic examination is a challenge current diagnostic practice. For first time we demonstrate use compression optical coherence elastography (C-OCE) technology as new functional OCT modality differentiating between cancerous and non-cancerous tissues colon detecting their morphological features on basis measurement tissue elastic properties. The method uses pre-determined stiffness values (Young's modulus) to distinguish different structures normal (mucosa submucosa), benign tumor (adenoma) malignant (including cells, gland-like structures, cribriform stromal fibers, extracellular mucin). After analyzing excess fifty samples, threshold value 520 kPa was suggested above which areas colorectal were detected invariably. A high Pearson correlation (r =0.98; p <0.05), negligible bias (0.22) by good agreement segmentation results C-OCE histological (reference standard) images demonstrated, indicating efficiency identify localization possibility perform biopsy. Furthermore, demonstrated ability differentiate subtypes - low-grade high-grade adenocarcinomas, mucinous adenocarcinoma, patterns. obtained ex vivo highlight prospects high-level malignancy detection. future endoscopic will allow sampling simultaneous rapid analysis heterogeneous morphology tumors.

Language: Английский

---

Human colorectal cancer tissue assessment using optical coherence tomography catheter and deep learning

Journal of Biophotonics, Journal Year: 2022, Volume and Issue: 15(6)

Published: Feb. 12, 2022

Optical coherence tomography (OCT) can differentiate normal colonic mucosa from neoplasia, potentially offering a new mechanism of endoscopic tissue assessment and biopsy targeting, with high optical resolution an imaging depth ~1 mm. Recent advances in convolutional neural networks (CNN) have enabled application ophthalmology, cardiology, gastroenterology malignancy detection sensitivity specificity. Here, we describe miniaturized OCT catheter residual network (ResNet)-based deep learning model manufactured trained to perform automatic image processing real-time diagnosis the images. The has outer diameter 3.8 mm, lateral ~7 μm, axial ~6 μm. A customized ResNet is utilized classify colorectal An area under receiver operating characteristic (ROC) curve (AUC) 0.975 achieved distinguish between cancerous

Language: Английский

---

Novel Pixelwise Co-Registered Hematoxylin-Eosin and Multiphoton Microscopy Image Dataset for Human Colon Lesion Diagnosis

Journal of Pathology Informatics, Journal Year: 2022, Volume and Issue: 13, P. 100012 - 100012

Published: Jan. 1, 2022

Colorectal cancer presents one of the most elevated incidences worldwide. Colonoscopy relies on histopathology analysis hematoxylin-eosin (H&E) images removed tissue. Novel techniques such as multi-photon microscopy (MPM) show promising results for performing real-time optical biopsies. However, clinicians are not used to this imaging modality and correlation between MPM H&E information is clear. The objective paper describe make publicly available an extensive dataset fully co-registered that allows research community analyze relationship histopathological effect semantic gap prevents from correctly diagnosing images. provides a scanned tissue at 10x resolution (0.5 µm/px) 50 samples lesions obtained by colonoscopies colectomies. Diagnostics capabilities TPF were compared. Additionally, tiles virtually stained into means deep-learning model. A panel 5 expert pathologists evaluated different modalities three classes (healthy, adenoma/hyperplastic, adenocarcinoma). Results showed performance over was 65% while virtual staining method achieved 90%. can provide appropriate colorectal without need staining. existing among needs be corrected.

Language: Английский

---

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

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: Aug. 17, 2022

Abstract 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 currently becoming more relevant in 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 tumour responses vivo. Here we present artificial intelligence (AI) approach analyze resultant data. In this initial study, show that AI can successfully classify SBRT-relevant clinical at multiple timepoints (t = 2–4 weeks) following irradiation (10 Gy and 30 cohorts) induced changes detected networks. Practicality obtained results, challenges associated modest number animals, their successful mitigation via augmented data approaches, advantages using deep learning methodologies, are discussed. Extension encouraging study AI-based time-series analysis outcome predictions finer level gradations envisioned.

Language: Английский

---

Optical coherence tomography and convolutional neural networks can differentiate colorectal liver metastases from liver parenchyma ex vivo

Journal of Cancer Research and Clinical Oncology, Journal Year: 2022, Volume and Issue: 149(7), P. 3575 - 3586

Published: Aug. 12, 2022

Optical coherence tomography (OCT) is an imaging technology based on low-coherence interferometry, which provides non-invasive, high-resolution cross-sectional images of biological tissues. A potential clinical application the intraoperative examination resection margins, as a real-time adjunct to histological examination. In this ex vivo study, we investigated ability OCT differentiate colorectal liver metastases (CRLM) from healthy parenchyma, when combined with convolutional neural networks (CNN).Between June and August 2020, consecutive adult patients undergoing elective resections for CRLM were included in study. Fresh specimens scanned vivo, before fixation formalin, using table-top device at 1310 nm wavelength. Scanned areas marked histologically examined. pre-trained CNN (Xception) was used match scans their corresponding diagnoses. To validate results, stratified k-fold cross-validation (CV) carried out.A total 26 (containing approx. 26,500 total) obtained 15 patients. Of these, 13 normal parenchyma CRLM. The distinguished F1-score 0.93 (0.03), sensitivity specificity 0.94 (0.04) (0.04), respectively.Optical can distinguish between great accuracy vivo. Further studies are needed improve upon these results develop diagnostic technologies, such scanning margins.

Language: Английский

---

Special Issue on “Machine Learning/Deep Learning in Medical Image Processing”

Applied Sciences, Journal Year: 2021, Volume and Issue: 11(23), P. 11483 - 11483

Published: Dec. 3, 2021

Many recent studies on medical image processing have involved the use of machine learning (ML) and deep (DL) [...]

Language: Английский

---

Optical coherence tomography combined with convolutional neural networks can differentiate between intrahepatic cholangiocarcinoma and liver parenchyma ex vivo

Journal of Cancer Research and Clinical Oncology, Journal Year: 2023, Volume and Issue: 149(10), P. 7877 - 7885

Published: April 12, 2023

Abstract Purpose Surgical resection with complete tumor excision (R0) provides the best chance of long-term survival for patients intrahepatic cholangiocarcinoma (iCCA). A non-invasive imaging technology, which could provide quick intraoperative assessment margins, as an adjunct to histological examination, is optical coherence tomography (OCT). In this study, we investigated ability OCT combined convolutional neural networks (CNN), differentiate iCCA from normal liver parenchyma ex vivo. Methods Consecutive adult undergoing elective resections between June 2020 and April 2021 ( n = 11) were included in study. Areas interest specimens scanned vivo, before formalin fixation, using a table-top device at 1310 nm wavelength. Scanned areas marked histologically examined, providing diagnosis each scan. An Xception CNN was trained, validated, tested matching scans their corresponding diagnoses, through 5 × stratified cross-validation process. Results Twenty-four three-dimensional (corresponding approx. 85,603 individual) ten analysis. cross-validation, model achieved mean F1-score, sensitivity, specificity 0.94, 0.93, respectively. Conclusion Optical can Further studies are necessary expand on these results lead innovative vivo applications, such or endoscopic scanning.

Language: Английский

---