-
A Lightweight Algorithm for Classifying Ex Vivo Tissues Samples
Authors:
Tzu-Hao Li,
Ethan Murphy,
Allaire Doussan,
Ryan Halter,
Kofi Odame
Abstract:
In this paper, we present a novel algorithm for classifying ex vivo tissue that comprises multi-channel bioimpedance analysis and a hardware neural network. When implemented in a mixed-signal 180 nm CMOS process, the classifier has an estimated power budget of 39 mW and an area of 30 mm2. This means that the classifier can be integrated into the tip of a surgical margin assessment probe, for in vi…
▽ More
In this paper, we present a novel algorithm for classifying ex vivo tissue that comprises multi-channel bioimpedance analysis and a hardware neural network. When implemented in a mixed-signal 180 nm CMOS process, the classifier has an estimated power budget of 39 mW and an area of 30 mm2. This means that the classifier can be integrated into the tip of a surgical margin assessment probe, for in vivo use during radical prostatectomy. We tested our classifier on digital phantoms of prostate tissue and also on an animal model of ex vivo bovine tissue. The classifier achieved an accuracy of 90% on the prostate tissue phantoms, and an accuracy of 84% on the animal model.
△ Less
Submitted 26 June, 2024;
originally announced June 2024.
-
Analog Gated Recurrent Neural Network for Detecting Chewing Events
Authors:
Kofi Odame,
Maria Nyamukuru,
Mohsen Shahghasemi,
Shengjie Bi,
David Kotz
Abstract:
We present a novel gated recurrent neural network to detect when a person is chewing on food. We implemented the neural network as a custom analog integrated circuit in a 0.18 um CMOS technology. The neural network was trained on 6.4 hours of data collected from a contact microphone that was mounted on volunteers' mastoid bones. When tested on 1.6 hours of previously-unseen data, the neural networ…
▽ More
We present a novel gated recurrent neural network to detect when a person is chewing on food. We implemented the neural network as a custom analog integrated circuit in a 0.18 um CMOS technology. The neural network was trained on 6.4 hours of data collected from a contact microphone that was mounted on volunteers' mastoid bones. When tested on 1.6 hours of previously-unseen data, the neural network identified chewing events at a 24-second time resolution. It achieved a recall of 91% and an F1-score of 94% while consuming 1.1 uW of power. A system for detecting whole eating episodes -- like meals and snacks -- that is based on the novel analog neural network consumes an estimated 18.8uW of power.
△ Less
Submitted 1 August, 2022;
originally announced August 2022.
-
Extracting Fractional Inspiratory Time from Electrocardiograms
Authors:
Maria Nyamukuru,
Kofi Odame
Abstract:
Non-invasive at-home monitoring of lung and lung airways health enables the early detection and tracking of respiratory diseases like asthma and chronic obstructive pulmonary disease (COPD). Various proposed approaches estimate the respiratory rate and extract the respiratory waveform from an electrocardiogram (ECG) signal as a way to discreetly monitor lung health. Unfortunately, these approaches…
▽ More
Non-invasive at-home monitoring of lung and lung airways health enables the early detection and tracking of respiratory diseases like asthma and chronic obstructive pulmonary disease (COPD). Various proposed approaches estimate the respiratory rate and extract the respiratory waveform from an electrocardiogram (ECG) signal as a way to discreetly monitor lung health. Unfortunately, these approaches fail to accurately capture the respiratory cycle phase features, resulting in a non-specific, incomplete picture of lung health. This paper introduces an algorithm to extract more respiratory information from the ECG signal by framing the problem as a binary segmentation task. In addition to respiratory rate (RR), the algorithm derives the fractional inspiratory time (FIT), a direct measure of airway obstruction derived from respiratory phase information. The algorithm is based on a gated recurrent neural network that infers vital respiratory information from a single-lead ECG signal. We measure our algorithm's performance on 5 subjects from the MIMIC dataset and 5 subjects from the CEBS database. Our algorithm maintains exceptional performance in estimating the respiratory rate and outperforms current algorithms that extract the respiratory cycle phases and FIT/ inspiratory:expiratory ratio (IER). Our algorithm reports a root mean squared error (RMSE) of 0.06 in the computation of FIT (values range from 0.2-0.6) and a RMSE of 0.54 breaths per minute (bpm) for respiratory rate (values range from 8 - 28 breaths per minute (bpm)) on the MIMIC dataset, and an FIT RMSE of 0.11 and and RR RMSE of 0.66 bpm on the CEBS dataset.
△ Less
Submitted 10 October, 2021; v1 submitted 2 December, 2020;
originally announced December 2020.
-
Inferring COPD Severity from Tidal Breathing
Authors:
Kofi Odame,
Graham Atkins,
Maria Nyamukuru,
Katherine Fearon
Abstract:
Objective: To develop an algorithm that can infer the severity level of a COPD patient's airflow limitation from tidal breathing data that is collected by a wearable device.
Methods: Data was collected from 25 single visit adult volunteers with a confirmed or suspected diagnosis of chronic obstructive pulmonary disease (COPD). The ground truth airflow limitation severity of each subject was dete…
▽ More
Objective: To develop an algorithm that can infer the severity level of a COPD patient's airflow limitation from tidal breathing data that is collected by a wearable device.
Methods: Data was collected from 25 single visit adult volunteers with a confirmed or suspected diagnosis of chronic obstructive pulmonary disease (COPD). The ground truth airflow limitation severity of each subject was determined by applying the Global Initiative for Chronic Obstructive Lung Disease (GOLD) staging criteria to the subject's spirometry results. Spirometry was performed in a pulmonary function test laboratory under the supervision of trained clinical staff. Separately, the subjects' respiratory signal was measured during quiet breathing, and a classification model was built to infer the subjects' level of airflow limitation from the measured respiratory signal. The classification model was evaluated against the ground truth using leave-one-out testing.
Results: Severity of airway obstruction was classified as either mild/moderate or severe/very severe with an accuracy of 96.4%.
Conclusion: Tidal breathing parameters that are measured with a wearable device can be used to distinguish between different levels of airflow limitation in COPD patients.
△ Less
Submitted 24 April, 2021; v1 submitted 15 November, 2020;
originally announced November 2020.
-
Tiny Eats: Eating Detection on a Microcontroller
Authors:
Maria T. Nyamukuru,
Kofi M. Odame
Abstract:
There is a growing interest in low power highly efficient wearable devices for automatic dietary monitoring (ADM) [1]. The success of deep neural networks in audio event classification problems makes them ideal for this task. Deep neural networks are, however, not only computationally intensive and energy inefficient but also require a large amount of memory. To address these challenges, we propos…
▽ More
There is a growing interest in low power highly efficient wearable devices for automatic dietary monitoring (ADM) [1]. The success of deep neural networks in audio event classification problems makes them ideal for this task. Deep neural networks are, however, not only computationally intensive and energy inefficient but also require a large amount of memory. To address these challenges, we propose a shallow gated recurrent unit (GRU) architecture suitable for resource-constrained applications. This paper describes the implementation of the Tiny Eats GRU, a shallow GRU neural network, on a low power micro-controller, Arm Cortex M0+, to classify eating episodes. Tiny Eats GRU is a hybrid of the traditional GRU [2] and eGRU [3] to make it small and fast enough to fit on the Arm Cortex M0+ with comparable accuracy to the traditional GRU. The Tiny Eats GRU utilizes only 4% of the Arm Cortex M0+ memory and identifies eating or non-eating episodes with 6 ms latency and accuracy of 95.15%.
△ Less
Submitted 14 March, 2020;
originally announced March 2020.
-
An Optimized Recurrent Unit for Ultra-Low-Power Keyword Spotting
Authors:
Justice Amoh,
Kofi Odame
Abstract:
There is growing interest in being able to run neural networks on sensors, wearables and internet-of-things (IoT) devices. However, the computational demands of neural networks make them difficult to deploy on resource-constrained edge devices.
To meet this need, our work introduces a new recurrent unit architecture that is specifically adapted for on-device low power acoustic event detection (A…
▽ More
There is growing interest in being able to run neural networks on sensors, wearables and internet-of-things (IoT) devices. However, the computational demands of neural networks make them difficult to deploy on resource-constrained edge devices.
To meet this need, our work introduces a new recurrent unit architecture that is specifically adapted for on-device low power acoustic event detection (AED). The proposed architecture is based on the gated recurrent unit (`GRU') but features optimizations that make it implementable on ultra-low power micro-controllers such as the Arm Cortex M0+.
Our new architecture, the Embedded Gated Recurrent Unit (eGRU) is demonstrated to be highly efficient and suitable for short-duration AED and keyword spotting tasks. A single eGRU cell is 60x faster and 10x smaller than a GRU cell. Despite its optimizations, eGRU compares well with GRU across tasks of varying complexities.
The practicality of eGRU is investigated in a wearable acoustic event detection application. An eGRU model is implemented and tested on the Arm Cortex M0-based Atmel ATSAMD21E18 processor. The Arm M0+ implementation of the eGRU model compares favorably with a full precision GRU that is running on a workstation. The embedded eGRU model achieves a classification accuracy 95.3%, which is only 2% less than the full precision GRU.
△ Less
Submitted 13 February, 2019;
originally announced February 2019.
-
DeepCough: A Deep Convolutional Neural Network in A Wearable Cough Detection System
Authors:
Justice Amoh,
Kofi Odame
Abstract:
In this paper, we present a system that employs a wearable acoustic sensor and a deep convolutional neural network for detecting coughs. We evaluate the performance of our system on 14 healthy volunteers and compare it to that of other cough detection systems that have been reported in the literature. Experimental results show that our system achieves a classification sensitivity of 95.1% and a sp…
▽ More
In this paper, we present a system that employs a wearable acoustic sensor and a deep convolutional neural network for detecting coughs. We evaluate the performance of our system on 14 healthy volunteers and compare it to that of other cough detection systems that have been reported in the literature. Experimental results show that our system achieves a classification sensitivity of 95.1% and a specificity of 99.5%.
△ Less
Submitted 8 September, 2015;
originally announced September 2015.
