Conventional medical imaging technologies for cancer diagnosis use fixed geometric configuration of the source and the detector to image the target. Our hypothesis is that the dynamic maneuver of source and detector geometry will lead to better performance of medical imaging devices. Interrogating a target in a three dimensional space requires cooperation and coordination between the source and detector positions. A dual-arm robot can be useful for this type of dynamic interrogation. The goal of this project is to develop a robotic bimodal dynamic imaging system for improved tumor characterization. The research question we are looking at "what is the suitable control/path planning method to perform dynamic interrogation?" The method will be tested on a bimodal dynamic imaging system implemented on a dual-arm robot Baxter.
Robotic Bimodal Dynamic Imaging System Development
Current Research Projects
Smartphone-based Compression-Induced Scope Development
Often times doctors use clinical breast examination using palpation. The examination, however, provides a subjective measurement of embedded inclusion. Mimicking palpation into a device can lead to an objective measurement of mechanical properties of inclusion. Size, hardness, and mobility some key mechanical properties to indicate malignancy of tumor. Tactile imaging system can emulate touch sensation of human finger using an optical method. This system can be made more portable using a smartphone. This idea initiated the development of a smartphone-based compression-induced scope, which will be a good screening device for rural and remote regions. This scope can capture the images of inclusions, sent them over the cloud securely to a computing server, and display the risk score of embedded inclusion. Currently, we are working on implementing compression-induced scope on a smartphone and establish a framework for computing risk score of screened target.
Bimodal Dynamic Imaging System for Breast Cancer Diagnosis
Can multimodality improve the breast cancer diagnosis performance? Will dynamic positioning of source and detector improve tumor characterization performance? In this work, we attempted to answer these questions by building a bimodal dynamic imaging system. The bimodal imaging system includes tactile and spectral mode for mechanical and spectral characterization of the target. The mechanical and spectral properties are key factors to classify healthy tissue versus benign and malignant tumor. We developed a prototype of bimodal imaging system by using a tactile imaging system, a two-axis gimbal and a two-axis linear stage. Experiments showed that the bimodal imaging system performs better in tumor characterization compared to the tactile imaging system.
Hybrid Hierarchical Statistical Control Scheme Development
In this work, we present a hybrid hierarchical statistical control approach for robotic manipulators. For the bimodal dynamic imaging system, these robotic manipulators are utilized to move the source and the detector. The bimodal dynamic imaging system contains both the continuous and discrete dynamics. Therefore, only statistical control is not sufficient to model the system dynamics. We consider the bimodal dynamic imaging system as a hybrid system. Here, first we utilized a full-state feedback statistical controller
to minimize the joint angle variations of the previously developed single robot manipulator model of Baxter. Then, we considered two such statistical-controlled manipulators as agents and developed a supervisory controller for coordination between the manipulators. The feasibility of the hybrid hierarchical statistical controller is demonstrated with numerical simulations.
Statistical Stackelberg Game Control Development
In this work, we derive the Stackelberg solution of a two-player, nonzero-sum game, where the leader and the follower optimizes the system performance by shaping the n-th cumulant of their cost functions, while the follower minimizes the mean of a different cost function. We introduce the Stackelberg statistical game theory for the automatic control of the source and detector of the bimodal imaging system. The Stackelberg statistical game allows sequential and hierarchical moves for the players, while optimizing the n-th cumulant of each player’s cost function.
Tactile Imaging System for Screening Breast Cancer
Non-invasive mechanical property estimation of an embedded object (tumor) can be used in medicine for characterization between malignant and benign lesions. We developed a tactile imaging sensor which is capable of detecting mechanical properties of inclusions. Studies show that stiffness of tumor is a key physiological discerning parameter for malignancy. As our sensor compresses the tumor from the surface, the sensing probe deforms, and the
light scatters. This forms the tactile image. Using the features of the image, we can estimate the mechanical properties such as size, depth, and elasticity of the embedded object. To test the performance of the method, a phantom study was performed. Silicone rubber balls were used as embedded objects inside the tissue mimicking substrate made of Polydimethylsiloxane. The average relative errors for size, depth, and elasticity were found to be 67.5%, 48.2%, and 69.1%, respectively. To test the feasibility of the sensor in estimating the elasticity of tumor, a pilot clinical study was performed on twenty breast cancer patients. The estimated elasticity was correlated with the biopsy results. Preliminary results show that the sensitivity of 67% and the specificity of 91.7% for elasticity. Results from the clinical study suggest that the tactile imaging sensor may be used as a tumor malignancy characterization tool.
Cyber Physical System Framework for Tactile Imaging System
Currently biopsy is the gold standard of detecting malignant tumors. Only larger hospitals are equipped with complex tumor detection instruments (mammogram, MRI, CT, and ultrasound) and trained operators. Young women and women in the remote regions are reluctant to screen for breast cancer. In developing countries, even the hospitals lack breast cancer detecting instruments due to the cost. A simple, reliable, harmless (no radiation), and cost-effective malignant tumor detecting device that is available near the patient (e.g., primary care physicians’ offices) would significantly increase the early detection rate. A Cyber-Physical Tactile Imaging System (CP-TIS) can estimate the mechanical properties of the tumor and differentiate between malignant and benign tumors. Physical part consists of targets, TIS’s, computers, and operators. Cyber part consists of computation units, communication network, and database software. From the processed images, mechanical properties of the tumor such as size and Young’s Modulus will be determined. Based on the mechanical properties, an index for malignancy will be generated. All these information will be available for the concerned parties to decide whether to seek further medical help.
Tactile and Hyperspectral Imaging for Canine Mammary Tumor Classification
The use of both tactile and hyperspectral imaging sensors, which exploit the mechanical and physiological changes in tissues, can significantly increase the performance in automatic identification of tumors with malignant histopathology. Tactile imaging measures the elastic modulus of a tumor, whereas hyperspectral imaging detects important biochemical markers. Spontaneous mammary tumors in canines were used to demonstrate the efficacy of our approach. The tactile sensor achieved a sensitivity of 50% and a specificity of 100% in identifying malignant tumors. The sensitivity and specificity of the hyperspectral sensor were 71% and 76%, respectively. We investigated several machine learning techniques for fusing the tactile and spectral data, which increased the sensitivity and specificity to 86% and 97%, respectively. Our tactile and hyperspectral imaging sensors are noninvasive and harmless (no ionized radiation is used). These imaging sensors may not only eliminate unnecessary surgeries, but will also motivate the development of similar sensors for human clinical use, due to the fact that canine and human tumors have similar physiology and biology.
Adaptive Neural Control Architecture within a Resilient Control Framework
In this work, an Adaptive Neural Control (ANC) architecture is used for system replication and control within a Resilient Control framework. A dynamic model is chosen for our plant and a “maliciously attacked” plant. A Model Reference Adaptive Control (MRAC) architecture is used to replicate and control the plant to match an ideal reference system. At certain time, we replicate a malicious attack by changing plant parameters, injecting false data, or altering sensor data. This attacked plant is then replicated and controlled to match the reference system. Simulations were carried out to show that accurate system replication and resilient control is possible using adaptive neural networks.
Color Performance Assessment of Whole-Slide Imaging Device
Color reproducibility is an essential factor when evaluating WSI devices for determining substantial equivalence. Color truth is required to assess color reproducibility. Color truth of biological tissues is difficult to measure due to their microscopic structures. Artificial color test targets were used. However, they differ from biological tissues in spectral and structural characteristics and therefore might confound evaluation results. In this work, the color reproducibility of two whole-slide imaging (WSI) devices was evaluated with real tissue slides. Three tissue slides – human colon, skin, and kidney – were used to test a modern and a legacy WSI devices. The color truth of the tissue slides was obtained using a multispectral imaging system. The output WSI images were compared with the color truth to calculate the color difference for each pixel. A psychophysical experiment was also conducted to measure the perceptual color reproducibility (PCR) of the same slides with 4 subjects. The results show that the mean color differences of the modern, legacy, and monochrome devices are 10.94, 22.35, and 42.74 ΔE00, while their mean PCRs are 70.35%, 23.06%, and 0.91%, respectively. The regulatory relevance was to demonstrate quantitative methods for comparing color performance of a 510(K) whole-slide imaging submission against its predicate device
Design and Assessment of a Virtual Teaching Assistant for an Open Laboratory
In order to provide an on-demand, open electrical engineering laboratory, we developed an innovative software-based Virtual Open Laboratory Teaching Assistant (VOLTA). This web-based virtual assistant provides laboratory instructions, equipment usage videos, circuit simulation assistance, and hardware implementation diagnostics. VOLTA allows students to perform laboratory experiments anywhere at their convenience. A series of studies was carried out in an experimental section of our traditional entry-level electric circuits laboratory to answer the hypothesis: students taught using VOLTA will learn as much as students who were taught by a human teaching assistant? The experiments were conducted using a pre-test/post-test design, and student performance was assessed using ANOVA on gain scores. Statistical analysis revealed that the student performance increased significantly when VOLTA was integrated into a closed laboratory. We conclude that VOLTA can support students in open and closed laboratories as effectively as a human teaching assistant.
Past Research Projects
Output Feedback Minimum Cost Variance Control of a Laser Target System
The space-based solar power system is an alternative to the ground based solar power system because of its round-the-clock availability. For the space-based solar power transmission, the accurate pointing of a laser from space to ground poses a challenging control task. A gimbaled laser target system, which is used for pointing laser to a target, is a test bench for such a transmission system. The objective of this research is to determine the optimal controller for the gimbaled laser target system in terms of pointing error and error variation. In order to achieve the objective, we modeled the gimbaled laser target system, simulated the model with the controllers, and tested them on the test bench. We used a Proportional- Integral-Derivative (PID) controller as the basis for the performance comparison since it is the most commonly used control method in the industry. Then we compared the PID controller with two statistical control methods Linear Quadratic Gaussian (LQG), and Minimal Cost Variance (MCV) optimal controllers. Experimental results indicate that the statistical controllers will provide a design parameter either to improve the mean pointing error or the standard deviation of the pointing error for the gimbaled laser target system. Subsequently, we believe that the statistical controllers will improve the space-based solar power transmission efficiency.