-
Validating polyp and instrument segmentation methods in colonoscopy through Medico 2020 and MedAI 2021 Challenges
Authors:
Debesh Jha,
Vanshali Sharma,
Debapriya Banik,
Debayan Bhattacharya,
Kaushiki Roy,
Steven A. Hicks,
Nikhil Kumar Tomar,
Vajira Thambawita,
Adrian Krenzer,
Ge-Peng Ji,
Sahadev Poudel,
George Batchkala,
Saruar Alam,
Awadelrahman M. A. Ahmed,
Quoc-Huy Trinh,
Zeshan Khan,
Tien-Phat Nguyen,
Shruti Shrestha,
Sabari Nathan,
Jeonghwan Gwak,
Ritika K. Jha,
Zheyuan Zhang,
Alexander Schlaefer,
Debotosh Bhattacharjee,
M. K. Bhuyan
, et al. (8 additional authors not shown)
Abstract:
Automatic analysis of colonoscopy images has been an active field of research motivated by the importance of early detection of precancerous polyps. However, detecting polyps during the live examination can be challenging due to various factors such as variation of skills and experience among the endoscopists, lack of attentiveness, and fatigue leading to a high polyp miss-rate. Deep learning has…
▽ More
Automatic analysis of colonoscopy images has been an active field of research motivated by the importance of early detection of precancerous polyps. However, detecting polyps during the live examination can be challenging due to various factors such as variation of skills and experience among the endoscopists, lack of attentiveness, and fatigue leading to a high polyp miss-rate. Deep learning has emerged as a promising solution to this challenge as it can assist endoscopists in detecting and classifying overlooked polyps and abnormalities in real time. In addition to the algorithm's accuracy, transparency and interpretability are crucial to explaining the whys and hows of the algorithm's prediction. Further, most algorithms are developed in private data, closed source, or proprietary software, and methods lack reproducibility. Therefore, to promote the development of efficient and transparent methods, we have organized the "Medico automatic polyp segmentation (Medico 2020)" and "MedAI: Transparency in Medical Image Segmentation (MedAI 2021)" competitions. We present a comprehensive summary and analyze each contribution, highlight the strength of the best-performing methods, and discuss the possibility of clinical translations of such methods into the clinic. For the transparency task, a multi-disciplinary team, including expert gastroenterologists, accessed each submission and evaluated the team based on open-source practices, failure case analysis, ablation studies, usability and understandability of evaluations to gain a deeper understanding of the models' credibility for clinical deployment. Through the comprehensive analysis of the challenge, we not only highlight the advancements in polyp and surgical instrument segmentation but also encourage qualitative evaluation for building more transparent and understandable AI-based colonoscopy systems.
△ Less
Submitted 6 May, 2024; v1 submitted 30 July, 2023;
originally announced July 2023.
-
GastroVision: A Multi-class Endoscopy Image Dataset for Computer Aided Gastrointestinal Disease Detection
Authors:
Debesh Jha,
Vanshali Sharma,
Neethi Dasu,
Nikhil Kumar Tomar,
Steven Hicks,
M. K. Bhuyan,
Pradip K. Das,
Michael A. Riegler,
Pål Halvorsen,
Ulas Bagci,
Thomas de Lange
Abstract:
Integrating real-time artificial intelligence (AI) systems in clinical practices faces challenges such as scalability and acceptance. These challenges include data availability, biased outcomes, data quality, lack of transparency, and underperformance on unseen datasets from different distributions. The scarcity of large-scale, precisely labeled, and diverse datasets are the major challenge for cl…
▽ More
Integrating real-time artificial intelligence (AI) systems in clinical practices faces challenges such as scalability and acceptance. These challenges include data availability, biased outcomes, data quality, lack of transparency, and underperformance on unseen datasets from different distributions. The scarcity of large-scale, precisely labeled, and diverse datasets are the major challenge for clinical integration. This scarcity is also due to the legal restrictions and extensive manual efforts required for accurate annotations from clinicians. To address these challenges, we present \textit{GastroVision}, a multi-center open-access gastrointestinal (GI) endoscopy dataset that includes different anatomical landmarks, pathological abnormalities, polyp removal cases and normal findings (a total of 27 classes) from the GI tract. The dataset comprises 8,000 images acquired from Bærum Hospital in Norway and Karolinska University Hospital in Sweden and was annotated and verified by experienced GI endoscopists. Furthermore, we validate the significance of our dataset with extensive benchmarking based on the popular deep learning based baseline models. We believe our dataset can facilitate the development of AI-based algorithms for GI disease detection and classification. Our dataset is available at \url{https://osf.io/84e7f/}.
△ Less
Submitted 17 August, 2023; v1 submitted 16 July, 2023;
originally announced July 2023.
-
Can Adversarial Networks Make Uninformative Colonoscopy Video Frames Clinically Informative?
Authors:
Vanshali Sharma,
M. K. Bhuyan,
Pradip K. Das
Abstract:
Various artifacts, such as ghost colors, interlacing, and motion blur, hinder diagnosing colorectal cancer (CRC) from videos acquired during colonoscopy. The frames containing these artifacts are called uninformative frames and are present in large proportions in colonoscopy videos. To alleviate the impact of artifacts, we propose an adversarial network based framework to convert uninformative fra…
▽ More
Various artifacts, such as ghost colors, interlacing, and motion blur, hinder diagnosing colorectal cancer (CRC) from videos acquired during colonoscopy. The frames containing these artifacts are called uninformative frames and are present in large proportions in colonoscopy videos. To alleviate the impact of artifacts, we propose an adversarial network based framework to convert uninformative frames to clinically relevant frames. We examine the effectiveness of the proposed approach by evaluating the translated frames for polyp detection using YOLOv5. Preliminary results present improved detection performance along with elegant qualitative outcomes. We also examine the failure cases to determine the directions for future work.
△ Less
Submitted 4 April, 2023;
originally announced April 2023.
-
Accelerometric Method for Cuffless Continuous Blood Pressure Measurement
Authors:
Mousumi Das,
Tilendra Choudhary,
L. N. Sharma,
M. K. Bhuyan
Abstract:
Pulse transit time (PTT) has been widely used for cuffless blood pressure (BP) measurement. But, it requires more than one cardiovascular signals involving more than one sensing device. In this paper, we propose a method for continuous cuffless blood pressure measurement with the help of left ventricular ejection time (LVET). The LVET is estimated using a signal obtained through a micro-electromec…
▽ More
Pulse transit time (PTT) has been widely used for cuffless blood pressure (BP) measurement. But, it requires more than one cardiovascular signals involving more than one sensing device. In this paper, we propose a method for continuous cuffless blood pressure measurement with the help of left ventricular ejection time (LVET). The LVET is estimated using a signal obtained through a micro-electromechanical system (MEMS)-based accelerometric sensor. The sensor acquires a seismocardiogram (SCG) signal at the chest surface, and the LVET information is extracted. Both systolic blood pressure (SBP) and diastolic blood pressure (DBP) are estimated by calibrating the system with the original arterial blood pressure values of the subjects. The proposed method is evaluated using different quantitative measures on the signals collected from ten subjects under the supine position. The performance of the proposed method is also compared with two earlier approaches, where PTT intervals are estimated from electrocardiogram (ECG)-photoplethysmogram (PPG) and SCG-PPG, respectively. The performance results clearly show that the proposed method is comparable with the state-of-the-art methods. Also, the computed blood pressure is compared with the original one, measured through a CNAP system. It gives the mean errors of the estimated systolic BP and diastolic BP within the range of -0.19 +/- 3.3 mmHg and -1.29 +/- 2.6 mmHg, respectively. The mean absolute errors for systolic BP and diastolic BP are 3.2 mmHg and 2.6 mmHg, respectively. The accuracy of BPs estimated from the proposed method satisfies the requirements of the IEEE standard of 5 +/- 8 mmHg deviation, and thus, it may be used for ubiquitous long term blood pressure monitoring.
△ Less
Submitted 18 August, 2020;
originally announced August 2020.
-
Two-stream Fusion Model for Dynamic Hand Gesture Recognition using 3D-CNN and 2D-CNN Optical Flow guided Motion Template
Authors:
Debajit Sarma,
V. Kavyasree,
M. K. Bhuyan
Abstract:
The use of hand gestures can be a useful tool for many applications in the human-computer interaction community. In a broad range of areas hand gesture techniques can be applied specifically in sign language recognition, robotic surgery, etc. In the process of hand gesture recognition, proper detection, and tracking of the moving hand become challenging due to the varied shape and size of the hand…
▽ More
The use of hand gestures can be a useful tool for many applications in the human-computer interaction community. In a broad range of areas hand gesture techniques can be applied specifically in sign language recognition, robotic surgery, etc. In the process of hand gesture recognition, proper detection, and tracking of the moving hand become challenging due to the varied shape and size of the hand. Here the objective is to track the movement of the hand irrespective of the shape, size, and color of the hand. And, for this, a motion template guided by optical flow (OFMT) is proposed. OFMT is a compact representation of the motion information of a gesture encoded into a single image. In the experimentation, different datasets using bare hand with an open palm, and folded palm wearing green-glove are used, and in both cases, we could generate the OFMT images with equal precision. Recently, deep network-based techniques have shown impressive improvements as compared to conventional hand-crafted feature-based techniques. Moreover, in the literature, it is seen that the use of different streams with informative input data helps to increase the performance in the recognition accuracy. This work basically proposes a two-stream fusion model for hand gesture recognition and a compact yet efficient motion template based on optical flow. Specifically, the two-stream network consists of two layers: a 3D convolutional neural network (C3D) that takes gesture videos as input and a 2D-CNN that takes OFMT images as input. C3D has shown its efficiency in capturing spatio-temporal information of a video. Whereas OFMT helps to eliminate irrelevant gestures providing additional motion information. Though each stream can work independently, they are combined with a fusion scheme to boost the recognition results. We have shown the efficiency of the proposed two-stream network on two databases.
△ Less
Submitted 17 July, 2020;
originally announced July 2020.
-
Design of Breathing-states Detector for m-Health Platform using Seismocardiographic Signal
Authors:
Tilendra Choudhary,
L. N. Sharma,
M. K. Bhuyan,
Kangkana Bora
Abstract:
In this work, a seismocardiogram (SCG) based breathing-state measuring method is proposed for m-health applications. The aim of the proposed framework is to assess the human respiratory system by identifying degree-of-breathings, such as breathlessness, normal breathing, and long and labored breathing. For this, it is needed to measure cardiac-induced chest-wall vibrations, reflected in the SCG si…
▽ More
In this work, a seismocardiogram (SCG) based breathing-state measuring method is proposed for m-health applications. The aim of the proposed framework is to assess the human respiratory system by identifying degree-of-breathings, such as breathlessness, normal breathing, and long and labored breathing. For this, it is needed to measure cardiac-induced chest-wall vibrations, reflected in the SCG signal. Orthogonal subspace projection is employed to extract the SCG cycles with the help of a concurrent ECG signal. Subsequently, fifteen statistically significant morphological-features are extracted from each of the SCG cycles. These features can efficiently characterize physiological changes due to varying respiratory rates. Stacked autoencoder (SAE) based architecture is employed for the identification of different respiratory-effort levels. The performance of the proposed method is evaluated and compared with other standard classifiers for 1147 analyzed SCG-beats. The proposed method gives an overall average accuracy of 91.45% in recognizing three different breathing states. The quantitative analysis of the performance results clearly shows the effectiveness of the proposed framework. It may be employed in various healthcare applications, such as pre-screening medical sensors and IoT based remote health-monitoring systems.
△ Less
Submitted 5 April, 2021; v1 submitted 24 February, 2020;
originally announced February 2020.
-
Delineation and Analysis of Seismocardiographic Systole and Diastole Profiles
Authors:
Tilendra Choudhary,
M. K. Bhuyan,
L. N. Sharma
Abstract:
Precise estimation of fiducial points of a seismocardiogram (SCG) signal is a challenging problem for its clinical usage. Delineation techniques proposed in the existing literature do not estimate all the clinically significant points of an SCG signal, simultaneously. The aim of this research work is to propose a delineation framework to identify IM, AO, IC, AC, pAC and MO fiducial points with the…
▽ More
Precise estimation of fiducial points of a seismocardiogram (SCG) signal is a challenging problem for its clinical usage. Delineation techniques proposed in the existing literature do not estimate all the clinically significant points of an SCG signal, simultaneously. The aim of this research work is to propose a delineation framework to identify IM, AO, IC, AC, pAC and MO fiducial points with the help of a PPG signal. The proposed delineation method processes a wavelet-based scalographic PPG and an envelope construction scheme is proposed to estimate the prominent peaks of the PPG signal. A set of amplitude histogram based decision rules is developed for estimation of SCG diastole phases, namely AC, pAC and MO. Subsequently, the systolic phases, IM, AO and IC are detected by applying diastole masking on SCG and decision rules. Experimental results on real-time SCG signals acquired from our designed data acquisition-circuitry and their analysis show the effectiveness of the proposed scheme. Additionally, these estimated parameters are analyzed to show the discrimination between normal breathing and breathlessness conditions.
△ Less
Submitted 24 February, 2020;
originally announced February 2020.
