Armando Manduca Dept. of Physiology and Biophysics Mayo Clinic Rochester, MN 55905 PaulChristy Dept. of Diagnostic Radiology Mayo Clinic Rochester, MN 55905 Richard Ehman Dept. of Diagnostic Radiology Mayo Clinic Rochester, MN 55905 Abstract Avascular necrosis (AVN) of the femoral head is a common yet potentially seriousdisorder which can be detected in its very early stages with magnetic resonance imaging. We have developed multi-layer perceptron networks, trained with conjugate gradient optimization, which diagnose AVN from single magnetic resonance images of the femoral head with 100% accuracy on training data and 97% accuracy on test data. 1 INTRODUCTION Diagnostic radiology may be a very natural field of application for neural networks, since a simple answer is desired from a complex image, and the learning process that human experts undergo is to a large extent a supervised learning experience based on looking at large numbers of images with known interpretations. Although many workers have applied neural nets to various types of I-dimensional medical data (e.g. ECG and EEG waveforms), little work has been done on applying neural nets to diagnosis directly from medical images. We chose the diagnosis of avascular necrosis from magnetic resonance images as an ideal initial problem, because: the area in question is small and well-defined, its size and shape do not vary greatly between individuals, the condition (if present) is usually visible even at low spatial and gray level resolution on a single image, and real data is readily available.

Thesedevices are implanted and perform three types of actions: l.monitor the heart 2.to pace the heart 3.to apply high energy/high voltage electric shock 1bey sense the electrical activity of the heart through leads attached to the heart tissue. Two types of sensing are commooly used: Single Chamber: Lead attached to the Right Ventricular Apex (RVA) Dual Chamber: An additional lead is attached to the High Right Atrium (HRA). The actions performed by defibrillators are based on the outcome of a classification procedure based on the heart rhythms of different heart diseases (abnormal rhythms or "arrhythmias").

This work discusses various optimization techniques which were proposed in models for controlling arm movements. In particular, the minimum-muscle-tension-change model is investigated. A dynamic simulator of the monkey's arm, including seventeen single and double joint muscles, is utilized to generate horizontal hand movements. The hand trajectories produced by this algorithm are discussed.

A neurophysiologically-based model is presented that controls a simulated kinematic arm during goal-directed reaches. The network generates a quasi-feedforward motor command that is learned using training signals generated by corrective movements. For each target, the network selects and sets the output of a subset of pattern generators. During the movement, feedbackfrom proprioceptors turns off the pattern generators. The task facing individual pattern generators is to recognize when the arm reaches the target and to turn off. A distributed representation of the motor commandthat resembles population vectors seen in vivo was produced naturally by these simulations.

Accurate saccades require interaction between brainstem circuitry and the cerebeJJum. A model of this interaction is described, based on Kawato's principle of feedback-error-Iearning. In the model a part of the brainstem (the superior colliculus) acts as a simple feedback controJJer with no knowledge of initial eye position, and provides an error signal for the cerebeJJum to correct for eye-muscle nonIinearities. This teaches the cerebeJJum, modelled as a CMAC, to adjust appropriately the gain on the brainstem burst-generator's internal feedback loop and so alter the size of burst sent to the motoneurons. With direction-only errors the system rapidly learns to make accurate horizontal eye movements from any starting position, and adapts realistically to subsequent simulated eye-muscle weakening or displacement of the saccadic target.

In general, control objectives such as the positioning and orienting of the endeffector arespecified with respect to task space coordinates; however, the manipulator is typica1ly controlled only in the configuration space.

A large class of motor control tasks requires that on each cycle the controller istold its current state and must choose an action to achieve a specified, state-dependent, goal behaviour. This paper argues that the optimization of learning rate, the number of experimental control decisions beforeadequate performance is obtained, and robustness is of prime importance-if necessary at the expense of computation per control cycle andmemory requirement. This is motivated by the observation that a robot which requires two thousand learning steps to achieve adequate performance, or a robot which occasionally gets stuck while learning, will always be undesirable, whereas moderate computational expense can be accommodated by increasingly powerful computer hardware. It is not unreasonable toassume the existence of inexpensive 100 Mflop controllers within a few years and so even processes with control cycles in the low tens of milliseconds will have millions of machine instructions in which to make their decisions. This paper outlines a learning control scheme which aims to make effective use of such computational power. 1 MEMORY BASED LEARNING Memory-based learning is an approach applicable to both classification and function learningin which all experiences presented to the learning box are explicitly remembered. The memory, Mem, is a set of input-output pairs, Mem {(Xl, YI), (X21 Y2), ..., (Xb Yk)}.

A method for transforming performance evaluation signals distal both in space and time into proximal signals usable by supervised learning algorithms, presentedin [Jordan & Jacobs 90], is examined. A simple observation concerning differentiation through models trained with redundant inputs (as one of their networks is) explains a weakness in the original architecture and suggests a modification: an internal world model that encodes action-space exploration and, crucially, cancels input redundancy to the forward model is added. Learning time on an example task, cartpole balancing,is thereby reduced about 50 to 100 times. 1 INTRODUCTION In many learning control problems, the evaluation used to modify (and thus improve) controlmay not be available in terms of the controller's output: instead, it may be in terms of a spatial transformation of the controller's output variables (in which case we shall term it as being "distal in space"), or it may be available only several time steps into the future (termed as being "distal in time"). For example, control of a robot arm may be exerted in terms of joint angles, while evaluation may be in terms of the endpoint cartesian coordinates; furthermore, we may only wish to evaluate the endpoint coordinates reached after a certain period of time: the co- ·Current address: Computation and Neural Systems Program, California Institute of Technology, Pasadena CA. 563 564 Brody ordinatesreached at the end of some motion, for instance. In such cases, supervised learning methods are not directly applicable, and other techniques must be used. Here we study one such technique (proposed for cases where the evaluation is distal in both space and time by [Jordan & Jacobs 90)), analyse a source of its problems, and propose a simple solution for them which leads to fast, efficient learning. We first describe two methods, and then combine them into the "predictive forward modeling" technique with which we are concerned.

Many real-valued connectionist approaches to learning control realize exploration by randomness inaction selection. This might be disadvantageous when costs are assigned to "negative experiences" . The basic idea presented in this paper is to make an agent explore unknown regions in a more directed manner. This is achieved by a so-called competence map, which is trained to predict the controller's accuracy, and is used for guiding exploration. Based on this, a bistable system enables smoothly switching attention between two behaviors - exploration and exploitation - depending on expected costsand knowledge gain. The appropriateness of this method is demonstrated by a simple robot navigation task.

Following Simon's (1969) observation that complex behaviour may simply be the reflection of a complex environment a number of researchers (eg.