Ai-based alzheimer's disease detection and cognitive impairment level prediction method

Abstract

The present invention relates to an AI-based Alzheimer's disease detection and cognitive impairment level prediction method in which, by receiving acoustic features extracted from speech data and using a neural network, a mini-mental status examination (MMSE) score may be predicted, and Alzheimer's disease (AD) may be detected, that is, MMSE score prediction and AD detection may be simultaneously carried out using a single algorithm, and particularly, in order to increase the accuracy of the prediction and the detection, a loss function having both regression loss and classification loss reflected therein is used in the algorithm, and in order to enable the regression loss and the classification loss to act complementarily, a Bernoulli penalty is reflected in a mean squared error (MSE) function according to respective probability values of intervals, and the distance between a correct answer interval and a prediction interval is reflected as a weight in a cross entropy (CE) function, thereby enabling a predicted MMSE score to be distributed similarly to an actual MMSE score, and thus enabling MMSE score prediction and AD detection to be simultaneously carried out using a single algorithm model by using only acoustic features extracted from speech data.

Description

AI-based Alzheimer's diagnosis and cognitive impairment severity prediction method

[0001] The present invention relates to an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method, and more specifically, to an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method that can predict an MMSE (Mini-Mental Status Examination) score and diagnose Alzheimer's (AD) disease by using a neural network after receiving acoustic features extracted from voice data as input, that is, MMSE score prediction and Alzheimer's diagnosis can be performed simultaneously using a single algorithm, and in particular, in order to increase the accuracy of prediction and diagnosis, a loss function that reflects both regression loss and classification loss in the algorithm is utilized, and a Bernoulli penalty is reflected in the mean square error (MSE) function according to the probability value of each section so that the regression loss and the classification loss can work complementarily, and a cross entropy (CE) function reflects the distance between the correct section and the predicted section as a weight so that the predicted MMSE score can be distributed similarly to the actual MMSE score, thereby simultaneously performing MMSE score prediction and AD diagnosis with a single algorithm model using only acoustic features extracted from voice data.

[0002] Alzheimer's Disease (AD) is a neurodegenerative disease characterized by a long-term, progressive decline in cognitive function. It is a common form of dementia that initially progresses to mild cognitive impairment (MCI) and then to AD. Early detection of Alzheimer's disease is essential for timely treatment, which can slow cognitive decline and reduce the risk of developing dementia.

[0003] One of the most common methods for diagnosing AD is the Mini-Mental Status Examination (MMSE), a traditional test that measures cognitive function. The MMSE is inexpensive and allows for rapid diagnosis. The MMSE score ranges from 0 to 30, with a score of 27 or lower typically considered Alzheimer's to minimize misdiagnosis in patients with Mild Cognitive Impairment (MCI). Rather than simply assessing the presence of Alzheimer's, the MMSE can be used to assess the severity of cognitive function, allowing for monitoring the progression of Alzheimer's disease and serving as an indicator for long-term monitoring.

[0004] Typical symptoms of Alzheimer's disease include memory and cognitive decline, but language impairments are also common. Apraxia of speech, logopanic progressive aphasia (LPA), and anomia commonly occur. In more severe cases, patients may stutter, repeat thoughts and questions, and have difficulty forming simple sentences. In this regard, speech can be used to characterize the loss of verbal fluency associated with Alzheimer's disease, such as pauses and silences. Therefore, speech can serve as an important biomarker for rapid diagnosis of Alzheimer's disease and for monitoring cognitive status.

[0005] The task of diagnosing Alzheimer's disease from speech is divided into a regression task, predicting MMSE scores, and a classification task, determining whether or not the patient has Alzheimer's. To accomplish these two tasks, many studies have extracted features from speech and used them as model inputs. Typically, these features are extracted by distinguishing between acoustic and linguistic features. Acoustic features include eGeMAPS, MFCC, X-vector, and VGGish, while linguistic features include WER (Word Error Rate) and BERT embeddings.

[0006] Typically, models that use linguistic features rather than acoustic features have shown better performance. However, models that use linguistic features require features to be extracted by creating a transcription from speech manually or through automatic speech recognition (ASR). However, manually creating a transcription is expensive and time-consuming, and when using ASR, the model is vulnerable to noise and its performance is affected by ASR performance. Therefore, the need for a model that can accurately diagnose and predict AD using only acoustic features is increasing.

[0007] (Patent Document) Domestic Patent No. 10-2274072 (registered on July 1, 2021) "Method and Device for Determining the Degree of Dementia in a User"

[0008] The prior patent disclosed in the above <patent document> also only discloses the concept of visualizing the characteristics of the voice input by the user and using this to determine the degree of dementia.

[0009] In particular, existing prior technologies have the limitation that MMSE score prediction for predicting the degree of cognitive impairment and AD diagnosis for diagnosing Alzheimer's disease are performed using separate algorithms.

[0010] The present invention has been devised to solve the above problems.

[0011] The purpose of the present invention is to provide an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method that can predict MMSE (Mini-Mental Status Examination) scores and diagnose Alzheimer's Disease (AD) using a neural network by inputting acoustic features extracted from voice data, that is, can simultaneously perform MMSE score prediction and Alzheimer's diagnosis using a single algorithm.

[0012] Another object of the present invention is to provide an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method that utilizes a loss function that reflects both regression loss and classification loss in an algorithm to increase the accuracy of prediction and diagnosis.

[0013] Another object of the present invention is to provide an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method, in which the regression loss applies the mean squared error (MSE) function, the classification loss applies the cross-entropy (CE) function, and the mean squared error (MSE) function reflects a Bernoulli penalty according to the probability value of each section so that the regression loss and the classification loss can work complementarily, thereby causing a decrease in the mean squared error (MSE) to be linked to a decrease in the cross-entropy (CE).

[0014] Another object of the present invention is to provide an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method that can distribute predicted MMSE scores similarly to actual MMSE scores by reflecting the distance between the correct interval and the predicted interval as a weight in the cross entropy (CE) function.

[0015] Another object of the present invention is to provide an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method that can simultaneously perform MMSE score prediction and AD diagnosis using a single algorithm model using only acoustic features extracted from voice data.

[0016] In order to achieve the above-mentioned purpose, the present invention is implemented by an embodiment having the following configuration.

[0017] According to one embodiment of the present invention, the AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method according to the present invention is characterized by including a step (S1) of extracting acoustic features from voice data; a step (S3) of receiving the extracted acoustic features as input and predicting a Mini-Mental Status Examination (MMSE) score using a neural network and diagnosing Alzheimer's (AD; Alzheimer's Disease);

[0018] According to another embodiment of the present invention, the step S1 in the present invention is characterized by including a step (S11) of obtaining VGGish features by inputting a Mel-Spectrogram extracted from the voice data into a pre-learning model in order to reflect the context of the voice data; a step (S12) of performing global max pooling on the VGGish features for each voice segment; a step (S13) of applying dropout at a ratio of 0.1 to 0.3 to increase generalization performance; and a step (S14) of combining the VGGish features with a speaker ID.

[0019] According to another embodiment of the present invention, the step S3 in the present invention is characterized by using a complementary algorithm that simultaneously performs MMSE score prediction and Alzheimer's diagnosis.

[0020] According to another embodiment of the present invention, the MMSE score prediction in the present invention is characterized by dividing the MMSE score range into a plurality of equal ranges and predicting the probability of each range in order to optimize the predicted value to an actual distribution.

[0021] According to another embodiment of the present invention, the complementary algorithm in the present invention is characterized by utilizing a loss function that reflects both regression loss and classification loss to increase the accuracy of the predicted value.

[0022] According to another embodiment of the present invention, the regression loss is characterized in that a mean squared error (MSE) function is applied, and the classification loss is characterized in that a cross-entropy (CE) function is applied.

[0023] According to another embodiment of the present invention, the mean square error (MSE) function in the present invention is characterized in that a Bernoulli penalty is reflected according to the probability value of each section, so that a decrease in the mean square error (MSE) is linked to a decrease in cross entropy (CE).

[0024] According to another embodiment of the present invention, the cross entropy (CE) function of the present invention is characterized in that it reflects the distance between the correct interval and the predicted interval as a weight so as to reflect the degree of error between the correct interval and the predicted interval.

[0025] The present invention can obtain the following effects through the combination and use of the configuration described above and the following examples.

[0026] The present invention has the effect of being able to predict the MMSE (Mini-Mental Status Examination) score and diagnose Alzheimer's Disease (AD) by using a neural network by inputting acoustic features extracted from voice data, that is, it is possible to simultaneously perform MMSE score prediction and Alzheimer's diagnosis using a single algorithm.

[0027] The present invention has the effect of increasing the accuracy of prediction and diagnosis by utilizing a loss function that reflects both regression loss and classification loss in a single algorithm that simultaneously performs MMSE score prediction and Alzheimer's diagnosis.

[0028] The present invention applies a mean squared error (MSE) function to the regression loss, and applies a cross-entropy (CE) function to the classification loss, and reflects a Bernoulli penalty in the mean squared error (MSE) function according to the probability value of each section so that the regression loss and the classification loss can work complementarily, thereby having the effect of causing a decrease in the mean squared error (MSE) to be linked to a decrease in the cross-entropy (CE).

[0029] The present invention has the effect of distributing predicted MMSE scores similarly to actual MMSE scores by reflecting the distance between the correct interval and the predicted interval as a weight in the cross entropy (CE) function.

[0030] The present invention has the effect of simultaneously performing MMSE score prediction and AD diagnosis using a single algorithm model using only acoustic features extracted from voice data.

[0031] Figure 1 is a conceptual diagram of an algorithm model in which a method according to one embodiment of the present invention is implemented.

[0032] Figure 2 is a reference diagram illustrating the process of extracting and applying voice features from voice data.

[0033] Figure 3 is a reference diagram of a conventional MMSE score prediction model using the MSE function.

[0034] Figure 4 is a state diagram showing an example of dividing the MMSE score range into multiple equal ranges and predicting the probability of each range.

[0035] Figure 5 is a reference diagram explaining the MMSE test process using voice data in which the examinee describes the Cookie Theft photo.

[0036] Figure 6 shows the results reflecting Bernoulli penalty and distance-based weights in MMSE score prediction.

[0037] Figure 7 shows data comparing the performance of existing models that only use acoustic features in the ADReSSo dataset.

[0038] Hereinafter, preferred embodiments of the AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method according to the present invention will be described in detail with reference to the attached drawings. It should be noted that identical components are represented by the same reference numerals throughout the drawings, as much as possible. Unless otherwise defined, all terms in this specification have the same general meaning as those terms understood by a person skilled in the art to which the present invention pertains, and if there is a conflict with the meaning of a term used in this specification, the definitions used in this specification shall apply. Throughout the specification, when a part is said to "include" a certain component, this does not mean that other components are excluded, but rather that other components may be further included, unless specifically stated otherwise. In addition, terms such as "... part" and "... module" described in the specification mean a unit that processes at least one function or operation, and this may be implemented by hardware, software, or a combination of hardware and software.

[0039]

[0040] Referring to FIGS. 1 and 2, a method for diagnosing Alzheimer's disease and predicting the degree of cognitive impairment based on AI according to an embodiment of the present invention is capable of predicting a Mini-Mental Status Examination (MMSE) score and diagnosing Alzheimer's Disease (AD) by using a neural network after receiving acoustic features extracted from voice data as input, that is, simultaneously performing MMSE score prediction and Alzheimer's diagnosis using a single algorithm. To this end, the method is characterized by including a step (S1) of extracting acoustic features from voice data; a step (S3) of predicting a Mini-Mental Status Examination (MMSE) score and diagnosing Alzheimer's Disease (AD) by using a neural network after receiving the extracted acoustic features as input.

[0041] First, the S1 step is a process of extracting acoustic features from speech data. In particular, in the present invention, acoustic features are extracted from the entire speech data in order to reflect the context of the speech data. As explained above, in existing models for diagnosing Alzheimer's, models that use linguistic features rather than acoustic features have typically shown better performance. However, models that use linguistic features require features to be extracted by creating a transcription from speech manually or through automatic speech recognition (ASR). However, manually creating a transcription is costly and time-consuming, and when ASR is used, the model is vulnerable to noise and the performance of the model is affected by the performance of the ASR. Therefore, the present invention is characterized by being able to accurately diagnose AD and predict the degree of cognitive impairment using only acoustic features (sequential information of acoustic features).

[0042] A more specific description of the process of extracting acoustic features in the above step S1 is provided with reference to FIG. 2, which may include a step (S11) of inputting a Mel-Spectrogram extracted from speech data into a pre-training model to obtain VGGish features; a step (S12) of performing global max pooling on the VGGish features for each speech segment; a step (S13) of applying dropout at a ratio of 0.1 to 0.3 to improve generalization performance; and a step (S14) of combining the VGGish features with a speaker ID.

[0043] The above step S11 is a process of obtaining VGGish features by inputting the Mel-Spectrogram extracted from voice data into a pre-training model. The VGGish features can be obtained at 128 dimensions per second by inputting the Mel-Spectrogram extracted from voice into the pre-training model.

[0044] The above S12 step is a process of performing global max pooling on VGGish features for each voice segment. The extracted VGGish features are input to the model after performing global max pooling on each voice segment.

[0045] The above step S13 is a process of applying dropout at a ratio of 0.1 to 0.3 to improve generalization performance. In particular, it is preferable to apply dropout at a ratio of 0.2 to prevent overfitting in which the output value is influenced only by a specific feature and to improve generalization performance.

[0046] The above step S14 is the process of combining the VGGish features with the speaker ID. The VGGish features that went through dropout in the above step S13 are combined with the speaker ID after passing through the Attention Layer. The speaker ID is label-encoded for patients and nurses, and passes through a single Fully Connected (FC) layer to increase the dimensionality to 128, the same as the VGGish features, and is combined channel-wise with the VGGish features.

[0047] After the VGGish features are combined with the speaker ID, they sequentially pass through the SE Block, Bi-LSTM, and Classification Block, and are then used to predict the MMSE score through SoftMax. Each SE Block consists of three 1-d convolutions and Squeeze and Excitation (SE), and after passing through a total of six SE Blocks, they pass through a 3-layer Bi-LSTM. After passing through the Bi-LSTM, Global Max Pooling is performed on the bidirectional hidden state of the last layer of the Bi-LSTM, and then they are extracted to 512 dimensions and input to the Classification Block. In the Classification Block, three FC layers reduce the input dimensions to 128 dimensions, 32 dimensions, and finally B dimensions for the number of intervals, and then pass through SoftMax to finally predict the probability of each interval (which is used to predict the MMSE score).

[0048] The above S3 step is a process of receiving extracted acoustic features as input and using a neural network to predict the MMSE (Mini-Mental Status Examination) score and diagnose Alzheimer's Disease (AD). As described above, the S3 step of the present invention is characterized by simultaneously performing MMSE score prediction and Alzheimer's diagnosis using one algorithm (complementary algorithm).

[0049] That is, conventional models that predict MMSE scores or diagnose Alzheimer's disease in order to predict the degree of cognitive impairment using voice features have been separately performing MMSE score prediction and Alzheimer's disease diagnosis (AD Detection) using separate regression models and classification models, respectively. For example, the conventional MMSE score prediction model shown in FIG. 3 applies the MSE (Mean Squared Error) function to predict MMSE scores, but since the predicted value regresses around the mean, it is impossible to determine whether Alzheimer's disease is present with the predicted MMSE score, so MMSE score prediction and Alzheimer's disease diagnosis (AD Detection) still have to be performed using separate models.

[0050] In order to solve these problems, the present invention utilizes a uniquely designed complementary algorithm to simultaneously perform MMSE score prediction and Alzheimer's diagnosis. To this end, the complementary algorithm utilizes a loss function that reflects both regression loss and classification loss to increase the accuracy of the predicted value, as illustrated in Fig. 1.

[0051] Meanwhile, in order to optimize the predicted value to the actual distribution, the MMSE score prediction divides the MMSE score range into multiple equal sections, as in the example illustrated in Fig. 4, and predicts the probability of each section, and the MMSE score at this time is estimated as the expected value of the probability distribution. In the example illustrated in Fig. 4, the total score range of the MMSE (30 points) is divided into 10 equal sections of 3 points each, and the probability of each section is predicted, and then the MMSE score is estimated as the expected value of the probability distribution.

[0052] The complementary algorithm of the present invention is characterized in that the mean squared error (MSE) function is applied as the regression loss, and the cross-entropy (CE) function is applied as the classification loss. That is, in order to make the model's prediction similar to the actual correct answer distribution even with a small amount of data in predicting the MMSE score, the cross-entropy (CE) function used in classification is additionally reflected in addition to the mean squared error (MSE) function, which can be confirmed in the mathematical expression 4 below.

[0053]

[0054]

[0055]

[0056]

[0057] (In the above mathematical equations 1 to 4, is the MMSE score of the i-th patient within a batch of size N, is the probability of the jth interval of the ith MMSE score, is the probability of the jth interval of the ith predicted value)

[0058] That is, in the above mathematical expression 4, if the interval is well predicted through CE, the predicted MMSE score also becomes closer to the actual MMSE score, which has the characteristic of leading to a decrease in MSE.

[0059] However, in situations where there is little data, such as medical data, since MSE regresses the predicted value around the mean, a decrease in MSE does not necessarily lead to a decrease in CE, and thus a situation may arise where the two loss functions cannot complement each other. Therefore, the present invention additionally reflects a Bernoulli penalty in the mean squared error (MSE) function according to the probability value of each section so that a decrease in the mean squared error (MSE) is linked to a decrease in the cross entropy (CE) so that the regression loss and the classification loss can work complementarily even in situations where there is little data.

[0060] The modified MMSE calculation formula (MSE(M,g(P))) that reflects the Bernoulli penalty can be found in Equation 5 below.

[0061]

[0062] Because the variance of the Bernoulli distribution is highest when the probability is 0.5, if the predicted probability in each interval is similar to the remaining predicted probabilities, the MSE suffers a significant loss. This allows the model to predict the probability of each interval as close to 1 or 0.

[0063] Meanwhile, since the cross-entropy (CE) function is generally calculated only with the probability of the correct interval, it does not reflect the degree of error between the correct interval and the predicted interval. Therefore, this patent proposes Weighted CE, which reflects the distance between the correct interval and the predicted interval as a weight to reflect the degree of error between them. In other words, by reflecting the distance between the correct interval and the predicted interval as a weight in the cross-entropy (CE) function, the predicted MMSE score can be distributed similarly to the actual MMSE score.

[0064] The distance between the correct interval and the predicted interval is defined by two things: the absolute distance and the square distance, which can be confirmed through mathematical equations 6 and 7 below, and the proposed Weighted CE can be confirmed through mathematical equation 8 below.

[0065]

[0066]

[0067]

[0068] (At this time, and represent the absolute distance and the square distance, respectively)

[0069] Finally, the weighted MSE-CE loss function to which the Bernoulli penalty is applied in the present invention is as shown in Equation 9 below.

[0070]

[0071] In this way, the weighted MSE-CE loss function with Bernoulli penalty applied makes MSE and CE complement each other by applying Bernoulli penalty to the MMSE calculation formula, and by reflecting the distance between the correct answer interval and the predicted interval as a weight in CE, it is possible to learn to predict similarly to the actual correct answer distribution.

[0072] In the following experiments, we verify the effectiveness of the weighted MSE-CE loss function with the Bernoulli penalty proposed in the present invention, and compare its performance with existing models that use only acoustic features on the ADReSSo dataset (Koo et al.'s model that won first place in the ADReSSo Challenge 2020 (Koo and Junghyun, "Exploiting multi-modal features from pre-trained networks for Alzheimer's dementia recognition." in the Proc. of Interspeech, September 2020)).

[0073] [Experiment] Experimental Effects of a Weighted MSE-CE Loss Function with Bernoulli Penalty

[0074] - ADReSSo dataset

[0075] The ADReSSo dataset is designed for Alzheimer's disease classification, cognitive ability assessment score inference, and cognitive decline prediction. It consists of 166 training datasets and 71 validation datasets. The audio files contain the subject's voice describing the Cookie Theft photo (see Figure 5), and may include a nurse's voice. In this patent, VGGish features extracted from the audio files and speaker-specific voice segments are used as input for the model.

[0076] - Experimental method

[0077] The MMSE Score Prediction model is tested by setting the number of intervals B to 10 and the penalty tuning parameter λ to 1. AD Detection is predicted by thresholding the predicted value of the model that performed MMSE Score Prediction at a score of 27. Finally, all training data are augmented with time inversion for the experiment.

[0078] - Experimental results

[0079] Referring to Fig. 6, when only MSE was used as a loss function, as in the model of Koo et al. that won first place in the ADReSSo Challenge 2020, it showed an RMSE of 6.60, which is very low performance, and when AD Detection classification was performed with the predicted scores, it showed an accuracy of 49.30%, which is very low because the model mostly predicted scores around the mean.

[0080] At this time, if CE is additionally reflected in the loss function, as in the model proposed in the present invention (Jin et al. model), the model's predicted values ​​can be trained to be similar to the distribution of actual MMSE scores, and the reduction in classification loss for the interval leads to a reduction in MSE, and it can be seen that RMSE is reduced to 5.49. Through more accurate MMSE score prediction, it can be confirmed that the classification accuracy in AD Detection also increases by 16.90%.

[0081] Here, when the Bernoulli penalty is reflected in the MMSE calculation formula, as in the model (Proposal model) ultimately proposed in this invention, the decrease in MSE leads to a decrease in CE, and the two loss functions complement each other. Therefore, we can see that the RMSE improves by 0.70 and the accuracy improves by 4.22% compared to when the Bernoulli penalty is not applied. Additionally, when the weighted CE is used, the RMSE achieves 4.55 and the accuracy reaches 77.46%, which is the best performance when using the absolute distance. However, when only the weighted CE is reflected without the Bernoulli penalty, the performance is improved only in the MMSE score prediction (since the two loss functions do not complement each other).

[0082] - Comparison with a model that uses only acoustic features from the ADReSSo dataset.

[0083] Referring to Fig. 7, 1) Among the models that performed MMSE score prediction, Luz et al. [S. Luz, F. Haider, Sdl Fuente, D. Fromm, B. MacWhinney "Detecting cognitive decline using speech only: The address challenge." in Proc, Interspeech, 2021, pp. 3780-3784] performed SVM with eGeMAPS and obtained an RMSE of 6.09. P rez-Toro et al.[PA P rez-Toro, et al., "Influence of the Interviewer on the Automatic Assessment of Alzheimer's Disease in the Context of the ADReSSo Challenge," in Proc, Interspeech, 2021, pp. 3785-3789] achieved the best RMSE of 5.35 among models that used only existing acoustic features through linear regression using Prosody Feature, X-vector, and Valence and Dominance features of VAD. However, this patent (Weighted MSE-CE Loss with Bernoulli Penalty) achieves a significantly improved RMSE of 4.55 while simultaneously performing MMSE score prediction and AD Detection using only VGGish features. 2) In addition, among models that performed AD Detection, Luz et al. obtained an accuracy of 64.79% with eGeMAPS. Gauder et al.[L. Gauder, L. Pepino, L. Ferrer, P. Riera, "Alzheimer Disease Recognition Using Speech-Based Embeddings From Pre-Trained Models," in Proc. Interspeech, 2021, pp. 3795-3799] achieved the highest accuracy of 78.9% among models that use only acoustic features by stacking 1D CNN layers using Wav2Vec 2.0. However, in this patent, a higher accuracy of 80.28% was achieved while simultaneously performing MMSE score prediction and AD detection, and a similar accuracy of 77.46% was achieved without data augmentation. 3) Finally, in this patent, the highest performance was achieved in both MMSE score prediction and AD detection among models that use only acoustic features in the ADReSSo dataset.In particular, unlike existing models that performed the two tasks as separate models, this patent achieved higher performance while simultaneously performing both tasks as a single model.

[0084] In conclusion, the model proposed in this patent reflects both regression loss and classification loss in MMSE score prediction, and reflects a penalty for the predicted probability value and a distance-based weight on the classification loss. The Bernoulli penalty makes the regression loss and the classification loss complement each other during learning, and the distance-based weight allows the model to better learn the distribution of actual MMSE scores. As confirmed in the above experiment, it achieved the highest RMSE of 4.55 and accuracy of 80.28% among the results of MMSE score prediction and AD detection performed using only acoustic features in the ADReSSo dataset. In this way, the present invention provides an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method that can predict the MMSE (Mini-Mental Status Examination) score and diagnose Alzheimer's Disease (AD) using a neural network by inputting acoustic features extracted from voice data, that is, it can simultaneously perform MMSE score prediction and Alzheimer's diagnosis using a single algorithm, and at the same time, the regression loss and the classification loss can work complementarily, by reflecting the Bernoulli penalty according to the probability value of each section in the mean square error (MSE) function, and by reflecting the distance between the correct section and the predicted section as a weight in the cross entropy (CE) function, so that the predicted MMSE score can be distributed similarly to the actual MMSE score, thereby enabling more accurate prediction and diagnosis.

[0085]

[0086] In the above, the applicant has described various embodiments of the present invention, but such embodiments are only examples of implementing the technical idea of ​​the present invention, and any change or modification that implements the technical idea of ​​the present invention should be interpreted as falling within the scope of the present invention.

Claims

1. Step of extracting acoustic features from voice data (S1); A method for diagnosing Alzheimer's disease and predicting the degree of cognitive impairment based on AI using acoustic features, characterized by including a step (S3) of predicting a Mini-Mental Status Examination (MMSE) score and diagnosing Alzheimer's disease (AD) by using a neural network by inputting extracted acoustic features.

2. In paragraph 1, The above S1 step targets the entire speech data to reflect the context of the speech data. A method for diagnosing Alzheimer's disease and predicting the degree of cognitive impairment based on AI using acoustic features, characterized by comprising the steps of: (S11) obtaining VGGish features by inputting a Mel-Spectrogram extracted from voice data into a pre-training model; (S12) performing global max pooling on the VGGish features for each voice segment; (S13) applying dropout at a ratio of 0.1 to 0.3 to improve generalization performance; and (S14) combining the VGGish features with a speaker ID.

3. In paragraph 1, The above S3 step is an AI-based method for diagnosing Alzheimer's disease and predicting the degree of cognitive impairment using acoustic features, characterized in that it uses a complementary algorithm that simultaneously performs MMSE score prediction and Alzheimer's diagnosis.

4. In paragraph 3, The above MMSE score prediction is an AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method using acoustic features, characterized in that the MMSE score range is evenly divided into multiple ranges to optimize the predicted value to the actual distribution and the probability of each range is predicted.

5. In paragraph 4, The above complementary algorithm is an AI-based method for diagnosing Alzheimer's disease and predicting the degree of cognitive impairment using acoustic features, characterized in that it utilizes a loss function that reflects both regression loss and classification loss to increase the accuracy of the predicted value.

6. In paragraph 5, An AI-based method for diagnosing Alzheimer's disease and predicting the degree of cognitive impairment using acoustic features, characterized in that the regression loss applies a mean squared error (MSE) function, and the classification loss applies a cross-entropy (CE) function.

7. In paragraph 6, An AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method using acoustic features, characterized in that the mean square error (MSE) function reflects a Bernoulli penalty according to the probability value of each section, so that a decrease in the mean square error (MSE) is linked to a decrease in cross entropy (CE).

8. In paragraph 7, An AI-based Alzheimer's diagnosis and cognitive impairment degree prediction method using acoustic features, characterized in that the cross entropy (CE) function above reflects the distance between the correct answer interval and the predicted interval as a weight so as to reflect the degree of error between the correct answer interval and the predicted interval.

Images