Cross-narrative Temporal Ordering of Medical Events
Raghavan, Preethi and Fosler-Lussier, Eric and Elhadad, Noémie and Lai, Albert M.

Article Structure

Abstract

Cross-narrative temporal ordering of medical events is essential to the task of generating a comprehensive timeline over a patient’s history.

Introduction

Discourse structure, logical flow of sentences, and context play a large part in ordering medical events based on temporal relations within a clinical narrative.

Related Work

In the areas of summarization and text-to-text generation, there has been prior work on several ordering strategies to order pieces of information extracted from different input documents (Barzilay et al., 2002, Lapata, 2003, Bollegala et al., 2010).

Problem Description

Medical events are temporally-associated concepts in clinical text that describe a medical condition affecting the patient’s health, or procedures performed on a patient.

Topics

coreference

Appears in 44 sentences as: coref (2) corefer (7) Coreference (1) coreference (35) coreferences (5)
In Cross-narrative Temporal Ordering of Medical Events
  1. The cross-narrative coreference and temporal relation weights used in both these approaches are learned from a corpus of clinical narratives.
    Page 1, “Abstract”
  2. These cross-narrative coreferences act as important anchors for reasoning with information across narratives.
    Page 1, “Introduction”
  3. We leverage cross-narrative coreference information along with confident cross-narrative temporal relation predictions and learn to align and temporally order medical event sequences across longitudinal clinical narratives.
    Page 1, “Introduction”
  4. The cross-narrative coreference and temporal relation scores used in both these approaches are learned from a corpus of patient narratives from The Ohio State University Wexner Medical Center.
    Page 1, “Introduction”
  5. We use dynamic programming to compute the best alignment, given the temporal and coreference information between medical events across these sequences.
    Page 2, “Related Work”
  6. elstart 2 628mm; and elstop = 6287501,, when 61 and 62 corefer .
    Page 3, “Problem Description”
  7. Thus, in order to align event sequences, we need to compute scores corresponding to cross-narrative medical event coreference resolution and cross-narrative temporal relations.
    Page 3, “Problem Description”
  8. 4 Cross-Narrative Coreference Resolution and Temporal Relation Learning
    Page 3, “Problem Description”
  9. Thus, for use in our sequence alignment models, we learn two independent classifiers for medical event coreference and temporal relation learning across narratives.
    Page 4, “Problem Description”
  10. We train a classifier to resolve cross-narrative coreferences by extracting semantic and temporal relatedness feature sets for each pair of medical concepts.
    Page 4, “Problem Description”
  11. Extracting these feature sets helps us train a classifier to predict medical event coreferences (Raghavan et al., 2012a).
    Page 4, “Problem Description”

See all papers in Proc. ACL 2014 that mention coreference.

See all papers in Proc. ACL that mention coreference.

Back to top.

dynamic programming

Appears in 23 sentences as: Dynamic Programming (2) Dynamic programming (2) dynamic programming (19)
In Cross-narrative Temporal Ordering of Medical Events
  1. We address the problem of aligning multiple medical event sequences, corresponding to different clinical narratives, comparing the following approaches: (1) A novel weighted finite state transducer representation of medical event sequences that enables composition and search for decoding, and (2) Dynamic programming with iterative pairwise alignment of multiple sequences using global and local alignment algorithms.
    Page 1, “Abstract”
  2. We present results using both approaches and observe that the finite state transducer approach performs performs significantly better than the dynamic programming one by 6.8% for the problem of multiple-sequence alignment.
    Page 1, “Abstract”
  3. As a contrast, we adapt dynamic programming algorithms (Needleman et al., 1970, Smith and Waterman, 1981) used to produce global and local alignments for aligning sequences of medical events across narratives.
    Page 1, “Introduction”
  4. dynamic programming or other ILP-based methods proposed in literature.
    Page 2, “Introduction”
  5. Dynamic programming algorithms have been popularly leveraged to produce pairwise and global genetic alignments, where edit distance based metrics are used to compute the cost of insertions, deletions and substitutions.
    Page 2, “Related Work”
  6. We use dynamic programming to compute the best alignment, given the temporal and coreference information between medical events across these sequences.
    Page 2, “Related Work”
  7. We demonstrate that the WFST—based approach outperforms popularly used dynamic programming algorithms for multiple sequence alignment.
    Page 2, “Related Work”
  8. We propose a novel WFST—based representation that enables accurate decoding for MSA when compared to popularly used dynamic programming algorithms (Needleman et al., 1970, Smith and Waterman, 1981) or other state of the art methods (Do et al., 2012).
    Page 4, “Problem Description”
  9. These scores are used in both the WFST—based representation and decoding, as well as for dynamic programming .
    Page 5, “Problem Description”
  10. We also use popular dynamic programming algorithms (Needleman et al., 1970, Smith and Waterman, 1981) for sequence alignment of medical events across narratives and compare it to the WFST-based representation and decoding.
    Page 7, “Problem Description”
  11. 5.3 Pairwise Alignment using Dynamic Programming
    Page 7, “Problem Description”

See all papers in Proc. ACL 2014 that mention dynamic programming.

See all papers in Proc. ACL that mention dynamic programming.

Back to top.

highest scoring

Appears in 4 sentences as: highest scoring (5)
In Cross-narrative Temporal Ordering of Medical Events
  1. The best hypothesis corresponds to the highest scoring path which can be obtained using shortest path algorithms like Djik—stra’s algorithm.
    Page 6, “Problem Description”
  2. Backtracking starts at the highest scoring matrix cell and proceeds until a cell with score zero is encountered, yielding the highest scoring local alignment.
    Page 7, “Problem Description”
  3. For each patient and each method (WFST or dynamic programming), the output timeline to evaluate is the highest scoring candidate hypothesis derived as described above.
    Page 8, “Problem Description”
  4. OpenFST provides tools that can search for the highest scoring sequences accepted by the machine, and can sample from high-scoring sequences probabilistically, by treating the
    Page 8, “Problem Description”

See all papers in Proc. ACL 2014 that mention highest scoring.

See all papers in Proc. ACL that mention highest scoring.

Back to top.

beam search

Appears in 3 sentences as: beam search (3)
In Cross-narrative Temporal Ordering of Medical Events
  1. During composition we retain intermediate paths like M 33 utilizing the ability to do lazy composition (Mohri and Pereira, 1998) in order to facilitate beam search through the multi-alignment.
    Page 6, “Problem Description”
  2. However, performing a beam search over the composed WFST in equation 2 allows us to accommodate such constraints across multiple sequences.
    Page 7, “Problem Description”
  3. The accuracy of the WFST—based representation and beam search across all sequences using the coreference and temporal relation scores to obtain the combined aligned sequence is 78.9%.
    Page 8, “Problem Description”

See all papers in Proc. ACL 2014 that mention beam search.

See all papers in Proc. ACL that mention beam search.

Back to top.

coreference resolution

Appears in 3 sentences as: Coreference Resolution (1) coreference resolution (2)
In Cross-narrative Temporal Ordering of Medical Events
  1. Thus, in order to align event sequences, we need to compute scores corresponding to cross-narrative medical event coreference resolution and cross-narrative temporal relations.
    Page 3, “Problem Description”
  2. 4 Cross-Narrative Coreference Resolution and Temporal Relation Learning
    Page 3, “Problem Description”
  3. The coreference resolution performs with 71.5% precision and 82.3% recall.
    Page 8, “Problem Description”

See all papers in Proc. ACL 2014 that mention coreference resolution.

See all papers in Proc. ACL that mention coreference resolution.

Back to top.

ILP

Appears in 3 sentences as: ILP (3)
In Cross-narrative Temporal Ordering of Medical Events
  1. We also compare the proposed methods with an Integer Linear Programming ( ILP ) based method for timeline construction (Do et al., 2012).
    Page 1, “Introduction”
  2. Moreover, it also outperforms the integer linear programming ( ILP ) method for timeline construction proposed in (Do et al., 2012).
    Page 9, “Problem Description”
  3. We observe that in case of MSA, the optimal solution using ILP is still intractable as the number of constraints increases exponentially with the number of sequences.
    Page 9, “Problem Description”

See all papers in Proc. ACL 2014 that mention ILP.

See all papers in Proc. ACL that mention ILP.

Back to top.

iteratively

Appears in 3 sentences as: iteratively (3)
In Cross-narrative Temporal Ordering of Medical Events
  1. One solution to this problem is to do he alignment greedily pairwise, starting from the most recent medical event sequences, finding the test path, and iteratively moving on to the next equence, and proceeding until the oldest medial event sequence.
    Page 6, “Problem Description”
  2. Thus, for MSA using dynamic programming, we use a heuristic method where we combine pairwise alignments iteratively starting with the latest narrative and progressing towards the oldest narrative.
    Page 7, “Problem Description”
  3. Aligning pairwise iteratively gives us an overall average accuracy of 68.2% similar to dynamic programming.
    Page 9, “Problem Description”

See all papers in Proc. ACL 2014 that mention iteratively.

See all papers in Proc. ACL that mention iteratively.

Back to top.