Machine Learning Derived Blood Input for Dynamic PET Images of Rat Heart

TL;DR

This work addresses non-invasive quantification in dynamic rodent PET by estimating the model-corrected input function (MCIF) from image-derived inputs. It introduces an LSTM network that takes concatenated IDIF and myocardial time-activity curves as input, with semi-automated segmentation used to derive the IDIF, and employs midpoint interpolation to alleviate late-time data sparsity. The interpolated LSTM significantly improves MCIF estimation, achieving about a 56% reduction in mean squared error compared with non-interpolated approaches, while maintaining stable dynamic similarity as measured by DTW. The study demonstrates the feasibility of automated, generalizable MCIF prediction in preclinical rodent PET and outlines future work toward fully automated pipelines and larger, more diverse datasets.

Abstract

Dynamic FDG PET imaging study of n = 52 rats including 26 control Wistar-Kyoto (WKY) rats and 26 experimental spontaneously hypertensive rats (SHR) were performed using a Siemens microPET and Albira trimodal scanner longitudinally at 1, 2, 3, 5, 9, 12 and 18 months of age. A 15-parameter dual output model correcting for spill over contamination and partial volume effects with peak fitting cost functions was developed for simultaneous estimation of model corrected blood input function (MCIF) and kinetic rate constants for dynamic FDG PET images of rat heart in vivo. Major drawbacks of this model are its dependence on manual annotations for the Image Derived Input Function (IDIF) and manual determination of crucial model parameters to compute MCIF. To overcome these limitations, we performed semi-automated segmentation and then formulated a Long-Short-Term Memory (LSTM) cell network to train and predict MCIF in test data using a concatenation of IDIFs and myocardial inputs and compared them with reference-modeled MCIF. Thresholding along 2D plane slices with two thresholds, with T1 representing high-intensity myocardium, and T2 representing lower-intensity rings, was used to segment the area of the LV blood pool. The resultant IDIF and myocardial TACs were used to compute the corresponding reference (model) MCIF for all data sets. The segmented IDIF and the myocardium formed the input for the LSTM network. A k-fold cross validation structure with a 33:8:11 split and 5 folds was utilized to create the model and evaluate the performance of the LSTM network for all datasets. To overcome the sparseness of data as time steps increase, midpoint interpolation was utilized to increase the density of datapoints beyond time = 10 minutes. The model utilizing midpoint interpolation was able to achieve a 56.4% improvement over previous Mean Squared Error (MSE).

Figures (8)

• Figure 1: A typical input function for a specific rodent. Each point represents a short scan of the body. Notice that the intervals between scans are not uniformly distributed. Only the first ten minutes are shown for representation.
• Figure 2: 2D map showing segmentation points (blue) of a rodent heart. Image A shows the segmentation of the myocardial tissue. Image B shows the segmentation of the left ventricule blood pool.
• Figure 3: 2 graphs of IDIF and model MCIF curves for an example rodent. Left: base curves with no interpolation. Right: Curves with interpolation after t = 10 minutes.
• Figure 4: The model architecture used to predict MCIFs from a concatenation of IDIFs and myocardial inputs - a single layer of 1000 LSTM cells, then time-distributed over a dense layer to regress the predicted MCIF using an interpolated 30 frame time series as input.
• Figure 5: Loss curve for the convergence of the algorithm. The x-axis represents epochs and the y-axis represents MSE. Early stopping can be seen in the fact that the model cuts off after $\sim$375 epochs.