Carlos Arteta

Learning to detect cells using non-overlapping extremal regions

Cell detection in microscopy images is an important step in the automation of cell based-experiments. We propose a machine learning-based cell detection method applicable to different modalities. The method consists of three steps: first, a set of candidate cell-like regions is identified. Then, each candidate region is evaluated using a statistical model of the cell appearance. Finally, dynamic programming picks a set of non-overlapping regions that match the model. The cell model requires few images with simple dot annotation for training and can be learned within a structured SVM framework. In the reported experiments, state-of-the-art cell detection accuracy is achieved for H&E stained histology, fluorescence, and phase-contrast images.

Interactive Object Counting

Our objective is to count (and localize) object instances in an image interactively. We target the regime where individual object detectors do not work reliably due to crowding, or overlap, or size of the instances, and take the approach of estimating an object density.

Detecting overlapping instances in microscopy images using extremal region trees.

In many microscopy applications the images may contain both regions of low and high cell densities corresponding to different tissues or colonies at different stages of growth. This poses a challenge to most previously developed automated cell detection and counting methods, which are designed to handle either the low-density scenario (through cell detection) or the high-density scenario (through density estimation or texture analysis). The objective of this work is to detect all the instances of an object of interest in microscopy images. The instances may be partially overlapping and clustered. To this end we introduce a tree-structured discrete graphical model that is used to select and label a set of non-overlapping regions in the image by a global optimization of a classification score. Each region is labeled with the number of instances it contains - for example regions can be selected that contain two or three object instances, by defining separate classes for tuples of objects in the detection process. We show that this formulation can be learned within the structured output SVM framework and that the inference in such a model can be accomplished using dynamic programming on a tree structured region graph. Furthermore, the learning only requires weak annotations - a dot on each instance. The candidate regions for the selection are obtained as extremal region of a surface computed from the microscopy image, and we show that the performance of the model can be improved by considering a proxy problem for learning the surface that allows better selection of the extremal regions. Furthermore, we consider a number of variations for the loss function used in the structured output learning. The model is applied and evaluated over six quite disparate data sets of images covering: fluorescence microscopy, weak-fluorescence molecular images, phase contrast microscopy and histopathology images, and is shown to exceed the state of the art in performance.

Learning to Detect Partially Overlapping Instances

The objective of this work is to detect all instances of a class (such as cells or people) in an image. The instances may be partially overlapping and clustered, and hence quite challenging for traditional detectors, which aim at localizing individual instances. Our approach is to propose a set of candidate regions, and then select regions based on optimizing a global classification score, subject to the constraint that the selected regions are non-overlapping. Our novel contribution is to extend standard object detection by introducing separate classes for tuples of objects into the detection process. For example, our detector can pick a region containing two or three object instances, while assigning such region an appropriate label. We show that this formulation can be learned within the structured output SVM framework, and that the inference in such model can be accomplished using dynamic programming on a tree structured region graph. Furthermore, the learning only requires weak annotations - a dot on each instance. The improvement resulting from the addition of the capability to detect tuples of objects is demonstrated on quite disparate data sets: fluorescence microscopy images and UCSD pedestrians.

Counting in the Wild

In this paper we explore the scenario of learning to count multiple instances of objects from images that have been dot-annotated through crowdsourcing. Specifically, we work with a large and challenging image dataset of penguins in the wild, for which tens of thousands of volunteer annotators have placed dots on instances of penguins in tens of thousands of images. The dataset, introduced and released with this paper, shows such a high-degree of object occlusion and scale variation that individual object detection or simple counting-density estimation is not able to estimate the bird counts reliably.

Low-Cost Blood Pressure Monitor Device for Developing Countries

Taking the Blood Pressure (BP) with a traditional sphygmomanometer requires a trained user. In developed countries, patients who need to monitor their BP at home usually acquire an electronic BP device with an automatic inflate/deflate cycle that determines the BP through the oscillometric method. For patients in resource constrained regions automated BP measurement devices are scarce because supply channels are limited and relative costs are high. Consequently, routine screening for and monitoring of hypertension is not common place. In this project we aim to offer an alternative strategy to measure BP and Heart Rate (HR) in developing countries. Given that mobile phones are becoming increasingly available and affordable in these regions, we designed a system that comprises low-cost peripherals with minimal electronics, offloading the main processing to the phone. A simple pressure sensor passes information to the mobile phone and the oscillometric method is used to determine BP and HR. Data are then transmitted to a central medical record to reduce errors in time stamping and information loss.

TeleSpiro: A low-cost mobile spirometer for resource-limited settings

Chronic obstructive pulmonary disease (COPD), a disabling combination of emphysema and chronic bronchitis, relies on spirometric lung function measurements for clinical diagnosis and treatment. Because spirometers are unavailable in most of the developing world, this project developed a low cost point of care spirometer prototype for the mobile phone called the “TeleSpiro.” The key contributions of this work are the design of a novel repeat-use, sterilisable, low cost, phone-powered prototype meeting developing world user requirements. A differential pressure sensor, dual humidity/pressure sensor, microcontroller and USB hardware were mounted on a printed circuit board for measurement of air flow in a custom machine-lathed respiratory air flow tube. The embedded circuit electronics were programmed to transmit data to and receive power directly from either a computer or Android smartphone without the use of batteries. Software was written to filter and extract respiratory cycles from the digitised data. Differential pressure signals from Telespiro showed robust, reproducible responses to the delivery of physiologic lung volumes. The designed device satisfied the stringent design criteria of resource-limited settings and makes substantial inroads in providing evidence-based chronic respiratory disease management.

A scalable mHealth system for noncommunicable disease management

Barriers to effective screening and management of NCDs in resource-constrained regions include limited availability of trained personnel, access to affordable automatic medical devices, and longitudinal clinical data. We present an end-to-end mHealth system which takes advantage of the almost universal availability of smartphones in order to address these barriers in a scalable and affordable manner. Our system includes simple, low-cost (

Multi-Task Convolutional Neural Network for Patient Detection and Skin Segmentation in Continuous Non-Contact Vital Sign Monitoring

5-

Respiratory rate estimation from the oscillometric waveform obtained from a non-invasive cuff-based blood pressure device.

20) and open-source peripherals that allow a minimally trained person to collect high-quality medical data at the point-of-care through a standard smartphone; allows the reliable transmission of clinical data even in the case of high-latency network connections; stores data in a cloud-based system, making patient records accessible anywhere; and enables both crowdsourced diagnostics and generation of annotated data for the research and development of automatic decision support and risk assessment systems. We show examples of the different elements of the system tailored for the management of cardiovascular disease and chronic obstructive pulmonary disease, which includes prototypes of the low-cost peripherals. In a validation study (of 40 volunteers), our smartphone-based blood pressure (BP) monitor was shown to measure BP, heart rate and respiration rate with a mean-absolute-error of less than 5 units from the reference values for 80% of the measurements.