TWI703961B - Oral image analysis system and method - Google Patents
Oral image analysis system and method Download PDFInfo
- Publication number
- TWI703961B TWI703961B TW107123466A TW107123466A TWI703961B TW I703961 B TWI703961 B TW I703961B TW 107123466 A TW107123466 A TW 107123466A TW 107123466 A TW107123466 A TW 107123466A TW I703961 B TWI703961 B TW I703961B
- Authority
- TW
- Taiwan
- Prior art keywords
- image
- platform
- complex
- classifier
- oral
- Prior art date
Links
Images
Landscapes
- Measuring And Recording Apparatus For Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
Description
本發明係有關於一口腔影像分析技術,特別係有關於一種根據口腔之取像位置、複數分類器,以及複數權重值來分析口腔影像之口腔影像分析技術。 The present invention relates to an oral image analysis technology, and particularly relates to an oral image analysis technology that analyzes oral images based on the oral image location, plural classifiers, and plural weight values.
傳統的口腔癌檢測方式包括醫師目視診察、觸診、唾液檢測、組織染色檢查或刷取活檢技術與病理組織切片檢查等。雖然切片檢查為目前口腔癌檢驗的主要方式,但仍有其應用上的缺點。舉例來說,由於切片檢查屬於侵入式醫療行為,因此傷口容易有感染風險。此外,不同經驗的醫療從業人員執行檢測/切片可能會對診斷結果產生相當大的差異。 Traditional oral cancer detection methods include physician visual inspection, palpation, saliva detection, tissue staining examination or brush biopsy technique and pathological tissue biopsy. Although biopsy is currently the main method of oral cancer inspection, there are still shortcomings in its application. For example, because biopsy is an invasive medical practice, wounds are prone to risk of infection. In addition, testing/sectioning performed by medical practitioners with different experience may produce considerable differences in the diagnosis results.
因此,為了降低因人工操作上的差異所導致的診斷誤差,如何達到更精確以及更有效率之口腔影像分析方式係一值得研究之課題。 Therefore, in order to reduce the diagnostic error caused by the difference in manual operation, how to achieve a more accurate and efficient oral image analysis method is a topic worthy of research.
有鑑於上述先前技術之問題,本發明提供了一種根據口腔之取像位置、複數分類器,以及複數權重值來分析口腔影像之口腔影像分析系統和方法。 In view of the above-mentioned problems of the prior art, the present invention provides an oral image analysis system and method for analyzing oral images according to the oral image capturing position, the complex classifier, and the complex weight value.
根據本發明之一實施例提供了一種口腔影像分析系統。上述口腔影像分析系統包括一影像擷取裝置以及一第一平台。影像擷取裝置會產生一第一影像。第一平台包括一第一處理裝置,且會從上述影像擷取裝置取得上述第一影像。上述第一處理裝置會判斷上述第一影像對應到口腔之一位置,並根據上述位置,取得複數分類器和取得對應上述複數分類器之複數權重值。上述第一處理裝置會從上述第一影像中選取對應一可能病變區域之一第二影像,並根據上述複數分類器、上述複數權重值以及上述第二影像產生一分析結果。 According to an embodiment of the present invention, an oral image analysis system is provided. The aforementioned oral image analysis system includes an image capturing device and a first platform. The image capturing device generates a first image. The first platform includes a first processing device, and obtains the first image from the image capturing device. The first processing device determines that the first image corresponds to a position of the oral cavity, and obtains a complex classifier and a complex weight value corresponding to the complex classifier according to the position. The first processing device selects a second image corresponding to a possible lesion area from the first image, and generates an analysis result according to the complex classifier, the complex weight value, and the second image.
根據本發明的一實施例,上述口腔影像分析系統更包括一第二平台。上述第二平台包括一第二處理裝置,且會藉由一雲端網路從上述第一平台取得上述第一影像,或直接從上述第一平台取得上述第一影像。第二處理裝置會判斷上述第一影像對應到口腔之上述位置,並根據上述位置,取得上述複數分類器和取得對應上述複數分類器之複數權重值。第二處理裝置會從上述第一影像中選取對應上述可能病變區域之上述第二影像,並根據上述複數分類器、上述複數權重值,以及上述第二影像產生上述分析結果。 According to an embodiment of the present invention, the aforementioned oral image analysis system further includes a second platform. The second platform includes a second processing device, and obtains the first image from the first platform through a cloud network, or obtains the first image directly from the first platform. The second processing device determines that the first image corresponds to the position of the oral cavity, and obtains the complex classifier and the complex weight value corresponding to the complex classifier according to the position. The second processing device selects the second image corresponding to the possible lesion area from the first image, and generates the analysis result according to the complex classifier, the complex weight value, and the second image.
根據本發明之另一實施例,提供了一種口腔影像分析方法。上述口腔影像分析方法之步驟包括;藉由一影像擷取裝置產生一第一影像;藉由一平台從上述影像擷取裝置取得上述第一影像,其中上述平台係一第一平台或 一第二平台,且上述第二平台之運算能力優於上述第一平台之運算能力;判斷上述第一影像對應到口腔之一位置;根據上述位置,取得複數分類器和取得對應上述複數分類器之複數權重值;從上述第一影像中選取對應一可能病變區域之一第二影像;以及根據上述複數分類器、上述複數權重值以及上述第二影像產生一分析結果。 According to another embodiment of the present invention, an oral image analysis method is provided. The steps of the oral image analysis method include: generating a first image by an image capturing device; obtaining the first image from the image capturing device by a platform, wherein the platform is a first platform or a second platform Platform, and the computing power of the second platform is better than that of the first platform; judging that the first image corresponds to a position of the oral cavity; according to the position, obtaining a complex classifier and obtaining a complex weight corresponding to the complex classifier Value; select a second image corresponding to a possible lesion area from the first image; and generate an analysis result according to the complex classifier, the complex weight value, and the second image.
關於本發明其他附加的特徵與優點,此領域之熟習技術人士,在不脫離本發明之精神和範圍內,當可根據本案實施方法中所揭露之口腔影像分析系統和方法,做些許的更動與潤飾而得到。 With regard to other additional features and advantages of the present invention, those skilled in the art, without departing from the spirit and scope of the present invention, can make some changes and changes based on the oral image analysis system and method disclosed in the implementation method of this case. Retouched.
100‧‧‧口腔影像分析系統 100‧‧‧Oral Image Analysis System
110‧‧‧影像擷取裝置 110‧‧‧Image capture device
120‧‧‧第一平台 120‧‧‧First Platform
121、141‧‧‧處理裝置 121、141‧‧‧Processing device
122、142‧‧‧儲存裝置 122、142‧‧‧Storage device
123、143‧‧‧顯示裝置 123、143‧‧‧Display device
130‧‧‧雲端網路 130‧‧‧Cloud Network
140‧‧‧第二平台 140‧‧‧Second Platform
210、410‧‧‧部位判定模組 210, 410‧‧‧Part Judgment Module
220、420‧‧‧選取模組 220、420‧‧‧Select module
230、430‧‧‧運算模組 230、430‧‧‧Computer module
310‧‧‧權重決定器 310‧‧‧Weight Determinator
440‧‧‧優化模組 440‧‧‧Optimization Module
第1圖係根據本發明之一實施例所述之口腔影像分析系統100之方塊圖。 FIG. 1 is a block diagram of the oral
第2A圖係根據本發明之一實施例所述之第一平台120之方塊圖。 FIG. 2A is a block diagram of the
第2B圖係根據本發明之一實施例所述之處理裝置121之方塊圖。 FIG. 2B is a block diagram of the
第3A圖係根據本發明之一實施例所述之分類器之示意圖。 Figure 3A is a schematic diagram of a classifier according to an embodiment of the present invention.
第3B圖係根據本發明之一實施例所述之子分類器A進行分類之示意圖。 Figure 3B is a schematic diagram of classification according to sub-classifier A according to an embodiment of the present invention.
第4A圖係根據本發明之一實施例所述之第二平台140 之方塊圖。 FIG. 4A is a block diagram of the
第4B圖係根據本發明之一實施例所述之處理裝置141之方塊圖。 FIG. 4B is a block diagram of the
第5圖係本揭露之一實施例所述之口腔影像分析方法之流程圖。 Figure 5 is a flowchart of the oral image analysis method according to an embodiment of the present disclosure.
本章節所敘述的是實施本發明之一最佳方式,目的在於說明本發明之精神而非用以限定本發明之保護範圍,本發明之保護範圍當視後附之申請專利範圍所界定者為準。 The description in this chapter is one of the best ways to implement the present invention. The purpose is to explain the spirit of the present invention and not to limit the scope of protection of the present invention. The scope of protection of the present invention shall be defined as the appended patent quasi.
第1圖係根據本發明之一實施例所述之口腔影像分析系統100之方塊圖。如第1圖所示,口腔影像分析系統100中包括了一影像擷取裝置110、一第一平台120、一雲端網路130以及一第二平台140。需注意的是,在第1圖中之方塊圖,僅係為了方便說明本發明之實施例,但本發明並不以此為限。口腔影像分析系統100亦可包括其他元件。 FIG. 1 is a block diagram of the oral
在本發明之實施例中,第一平台120可視為一連接影像擷取裝置110之控制平台,第二平台140則可視為一遠端之資料庫及處理與分析平台。第一平台120和第二平台140都具有影像處理之能力,但第二平台140之運算能力會優於第一平台120之運算能力。因此,當臨床診斷時,使用者或操作者(例如:醫生)可利用影像擷取裝置110來擷取看診者之口腔影像,並根據第一平台120顯示 影像擷取裝置110所擷取之影像所產生之分析結果,對看診者之病況做初步之評估。當要進行病症之研究或討論時,使用者(例如:多位醫生和研究人員)可經由網路連上第二平台140,並根據第二平台140針對第一平台之收案結果所產生之分析結果,進行病症之評估以及研究。此外,根據本發明一實施例,第一平台120亦可僅用來將影像擷取裝置110所擷取之影像直接傳送給第二平台140,或透過雲端網路130傳送給第二平台140,其餘所有關於影像之分析之操作都交由第二平台140來進行。 In the embodiment of the present invention, the
根據本發明之一實施例,影像擷取裝置110可耦接(或電性連接)至第一平台120。根據本發明之一實施例,影像擷取裝置110可為一手持裝置。使用者(例如:醫生)藉由影像擷取裝置110針對所要取像之口腔之位置(部位),進行影像之擷取,以產生對應該位置之影像。 According to an embodiment of the present invention, the image capturing
根據本發明之一實施例,影像擷取裝置110包括一影像擷取模組(圖未顯示)、一光源模組(圖未顯示),以及一濾鏡切換模組(圖未顯示),但本發明並不以此為限。 According to an embodiment of the present invention, the image capturing
根據本發明之一實施例,影像擷取模組包括一鏡頭,其可用以拍攝口腔之特定位置(部位),以產生對應該位置之影像。根據本發明之一實施例,光源模組可提供不同波長之光源,例如:白光光源(自然光)、紫外光(Ultraviolet,UV)光源、或藍光光源等。由於人體本身就有許多光感應物質(Fluorochrome),受光照射時會放出不同的螢光,此即為自體螢光特性。利用口中異常組織和周 圍正常組織的自體螢光特性的不同,可鑑別出異常組織的存在。舉例來說,不同波長之光源可分別用來激發口腔細胞內之菸鹼醯胺腺嘌呤二核苷酸(NADH)、黃素腺嘌呤二核苷酸(FAD),以及膠原蛋白等產生螢光。由於在有腫瘤發生的情況下,細胞內組織成份,包括NADH的濃度、FAD的濃度、膠原蛋白、血液量、組織紋理等均會改變,因此將可利用螢光影像加以運算後以推測出所關注之口腔部位是否發生病變。根據本發明之一實施例,濾鏡切換模組配置在影像擷取模組和光源模組之間,且其可根據不同波長之光源,切換不同濾鏡,擷取不同波段之螢光影像。 According to an embodiment of the present invention, the image capturing module includes a lens, which can be used to capture a specific position (position) of the oral cavity to generate an image corresponding to the position. According to an embodiment of the present invention, the light source module can provide light sources of different wavelengths, such as white light (natural light), ultraviolet (Ultraviolet, UV) light, or blue light. Since the human body has many light-sensitive substances (Fluorochrome), it will emit different fluorescence when irradiated by light, which is the self-fluorescence characteristic. The difference in autofluorescence characteristics between abnormal tissues in the mouth and surrounding normal tissues can be used to identify the presence of abnormal tissues. For example, light sources of different wavelengths can be used to excite nicotine amide adenine dinucleotide (NADH), flavin adenine dinucleotide (FAD), and collagen to produce fluorescence in oral cells. . In the case of tumors, the intracellular tissue components, including the concentration of NADH, the concentration of FAD, collagen, blood volume, tissue texture, etc., will change, so fluorescent images can be used to calculate and infer the concern Whether there is any disease in the oral cavity. According to an embodiment of the present invention, the filter switching module is arranged between the image capturing module and the light source module, and it can switch different filters according to light sources of different wavelengths to capture fluorescent images of different wavelengths.
第2A圖係根據本發明之一實施例所述之第一平台120之方塊圖。如第2A圖所示,第一平台120可包括一處理裝置121、一儲存裝置122,以及一顯示裝置123。需注意的是,在第2A圖中之方塊圖,僅係為了方便說明本發明之實施例,但本發明並不以此為限。第一平台120亦可包括其他元件。 FIG. 2A is a block diagram of the
根據本發明之一實施例,處理裝置121可係一包含一或複數處理器之裝置。根據本發明之一實施例,儲存裝置122可用以儲存複數分類器和複數權重值。根據本發明一實施例,儲存裝置122所儲存之複數分類器和複數權重值等資料可預先經過分析和訓練所產生。第一平台120可將儲存裝置122儲存之複數分類器和複數權重值經由雲端網路130提供給第二平台140。根據本發明另一實施例,複數分類器和複數權重值亦可由第二平台提供給第一平台 120。根據本發明之一實施例,顯示裝置123可例如是一液晶螢幕或一觸控面板,但本發明不以此為限。 According to an embodiment of the present invention, the
在本發明之實施例所述之複數分類器可根據影像強度(intensity)運算、異質性(heterogeneity)演算法、強度差異向量(intensity difference vector,IDV)演算法、賈伯濾波器演算法(Gabor Filter)、小波轉換演算法(Wavelet Transform)以及希爾伯特黃轉換演算法(Hilbert Huang Transform)等演算法之一或多者,對預先收集之複數口腔影像進行計算所產生。也就是說,複數分類器可分別對應不同的影像特徵及演算法。根據上述演算法之一或多者,可用以產生每一分類器所對應之影像特徵。此外,根據本發明一實施例,每一分類器更可根據底下的分類器運算演算法再去做分類,例如支援向量機(support vector machine)、KNN分類器、機率型類神經網路(probabilistic neural network)、旋積類神經網路(convolutional neural networks CNN)、決策樹等,底下將會以第3A-3B圖為例來做說明。此外,在本發明之實施例中,口腔之不同位置對應到相同分類器會有不同的權重值。 The complex classifier described in the embodiment of the present invention can be based on image intensity (intensity) calculation, heterogeneity (heterogeneity) algorithm, intensity difference vector (intensity difference vector, IDV) algorithm, Gabor filter algorithm (Gabor Filter), Wavelet Transform (Wavelet Transform), and Hilbert Huang Transform (Hilbert Huang Transform) and other algorithms are one or more of which are generated by calculating the plural oral images collected in advance. In other words, the complex classifier can correspond to different image features and algorithms. According to one or more of the above algorithms, it can be used to generate the image features corresponding to each classifier. In addition, according to an embodiment of the present invention, each classifier can be further classified according to the underlying classifier calculation algorithm, such as support vector machine (support vector machine), KNN classifier, and probabilistic neural network (probabilistic neural network). neural network), convolutional neural networks (CNN), decision trees, etc., the following will take Figure 3A-3B as an example for illustration. In addition, in the embodiment of the present invention, different positions of the oral cavity corresponding to the same classifier will have different weight values.
根據本發明一實施例,可使用一機器學習方法來產生不同分類器之權重值。上述機器學習方法例如是自適應增強(AdaBoost)演算法或整體學習(ensemble)演算法,但本發明不以此為限。 According to an embodiment of the present invention, a machine learning method can be used to generate the weight values of different classifiers. The above-mentioned machine learning method is, for example, an adaptive boost (AdaBoost) algorithm or an ensemble algorithm, but the present invention is not limited thereto.
自適應增強(AdaBoost)演算法的概念是收集多個分類器,聚合成新的分類器,效果可以比原來這些分 類器還要強。它係利用陽春分類器之間互補的關係學習,在學習過程中調整分類器權重。根據本發明一實施例,處理裝置121更包括一權重決定器(gating),如第3A圖所示。權重決定器可根據影像資料測試各個分類器所得到之結果,來決定不同分類器之權重值。正確率較高之分類器給予較高之權重值,反之則給較低之權重值。 The concept of the adaptive enhancement (AdaBoost) algorithm is to collect multiple classifiers and aggregate them into new classifiers. The effect can be stronger than the original classifiers. It uses the complementary relationship learning between Yangchun classifiers to adjust the classifier weights during the learning process. According to an embodiment of the present invention, the
第2B圖係根據本發明之一實施例所述之處理裝置121之方塊圖。如第2B圖所示,處理裝置121可包括一部位判定模組210、一選取模組220以及一運算模組230。 FIG. 2B is a block diagram of the
根據本發明之一實施例,當第一平台120從影像擷取裝置110取得影像擷取裝置110所產生之影像(底下以第一影像稱之)後,部位判定模組210會根據第一影像,判斷第一影像所對應到口腔之位置(或部位),例如:兩頰、舌頭、上顎、舌頭下方、牙齦等不同口腔位置。也就是說,部位判定模組210會根據第一影像之特性,判斷第一平台120之使用者或操作者(例如:醫生)目前所取像之位置係口腔的哪一部位。 According to an embodiment of the present invention, when the
根據本發明之另一實施例,第一平台120之使用者或操作者(例如:醫生)亦可根據第一影像,直接經由第一平台120輸入(或選取)第一影像所對應到之部位。舉例來說,使用者可直接經由顯示裝置123或第一平台120之一操作介面(圖未顯示)輸入(或選取)第一影像所對應到之部位。 According to another embodiment of the present invention, the user or operator of the first platform 120 (for example, a doctor) can directly input (or select) the part corresponding to the first image through the
當部位判定模組210取得第一影像所對應到口腔之位置後,會將此資訊提供給運算模組230。根據本發明之一實施例,運算模組230根據此位置,會從儲存裝置122取得儲存裝置122所儲存之複數分類器,以及複數權重值。複數分類器對應口腔之每一位置會有不同的權重值。也就是說,相同分類器對應到不同口腔之位置可有不同的權重值。 After the
根據本發明之另一實施例,當部位判定模組210得知第一影像所對應到口腔之位置後,運算模組230會從儲存裝置122中所儲存之複數分類器取得對應該位置之一組分類器,以及取得對應該組分類器之複數權重值。在此實施例中,口腔之每一位置都會有其對應之一組分類器,以及對應該組分類器之複數權重值。需注意的是,相同分類器對應到不同口腔之位置可有不同的權重值。 According to another embodiment of the present invention, when the
根據本發明另一實施例,運算模組230亦可根據第一影像,直接計算出對應每一分類器之權重值。也就是說,在此實施例中,運算模組230會自行根據第一影像對應之口腔之位置及影像特徵,利用一機器學習方法來計算出對應每一分類器之權重值,上述機器學習方法例如是自適應增強(AdaBoost)演算法或整體學習(ensemble)演算法。 According to another embodiment of the present invention, the
根據本發明之一實施例,選取模組220可根據第一影像,從第一影像中選取對應一可能病變區域之一第二影像。具體來說,選取模組220可透過一影像辨識演算 法,判斷第一影像中可能發生病變之區域,並從第一影像中擷取出對應該可能病變區域之影像。 According to an embodiment of the present invention, the
根據本發明之另一實施例,第一平台120之使用者或操作者(例如:醫生)亦可根據第一影像,直接經由第一平台120輸入(或選取)出第一影像中可能發生病變之區域,以產生第二影像。舉例來說,使用者可直接經由顯示裝置123或第一平台120之一操作介面輸入(或選取)第一影像中可能發生病變之區域,以產生第二影像。 According to another embodiment of the present invention, the user or operator of the first platform 120 (for example, a doctor) can also directly input (or select) the disease that may occur in the first image based on the first image. Area to produce the second image. For example, the user can directly input (or select) an area in the first image that may have a lesion through one of the operating interfaces of the
當第二影像產生後,運算模組230會根據其所取得之複數分類器、複數權重值以及第二影像,產生一分析結果,並將分析結果傳送給顯示裝置123。顯示裝置123會顯示該分析結果,以供第一平台120之使用者或操作者(例如:醫生)參考。 After the second image is generated, the
第3A圖係根據本發明之一實施例所述複數分類器之示意圖。需注意的是,在第3A圖中之示意圖,僅係為了方便說明本發明之實施例,但本發明並不以此為限。如第3A圖所示,分類器A所採用之影像特徵係經由變異度演算法及影像強度演算法所產生之結果(即變異度和影像強度),分類器B所採用之影像特徵係經由小波轉換演算法所產生之結果(即小波係數)以及分類器C所採用之影像特徵係經由賈伯濾波器演算法和影像強度演算法所產生之結果(即賈伯係數和影像強度)。每一分類器更可根據不同分類器運算演算法去做分類。底下將以第3B圖為例做說明。 Figure 3A is a schematic diagram of the complex classifier according to an embodiment of the present invention. It should be noted that the schematic diagram in Figure 3A is only for the convenience of describing the embodiments of the present invention, but the present invention is not limited thereto. As shown in Figure 3A, the image features used by classifier A are the results (ie, variability and image intensity) produced by the variability algorithm and image intensity algorithm, and the image features used by classifier B are through wavelet The result produced by the conversion algorithm (ie, wavelet coefficients) and the image features used by classifier C are the results produced by the Jaber filter algorithm and the image intensity algorithm (ie, Jaber coefficients and image intensity). Each classifier can also be classified according to different classifier calculation algorithms. The following will take Figure 3B as an example.
第3B圖係根據本發明之一實施例所述之分類器A在進行分類之示意圖。如第3A圖所示,分類器A所採用之影像特徵係經由變異度演算法及影像強度演算法所產生之結果(即變異度和影像強度)。在此實施例中,運算模組230可利用一決策樹之方式將影像進行分類。如第3B圖所示,運算模組230會將影像之變異度大於30且影像強度高於40之影像判定為癌症。反之,若影像之變異度小於30,運算模組230會將影像之類別判定為正常,以及若影像之變異度大於30但影像強度低於40,運算模組230亦會將影像之類別判定為正常。此外,根據本發明另一實施例,運算模組230亦可使用支持向量機(Support Vector Machine.SVM)之方式實作分類器以進行影像分類,計算每一分類器所使用的影像特徵值,並且載入訓練好的邊界線,以位置來判定輸入的影像係屬於哪一個類別,其中邊界線是透過支持向量機之方式,根據過去收集的資料訓練所產生的,其邊界能夠分隔兩類之間最大距離。此外,根據本發明另一實施例,運算模組230亦可使用一機率型類神經網路將每一分類器再進行分類,根據影像與訓練資料相似度,決定影像在各分類之可能性。因此,經過再分類後之每一分類器之判定結果可為疾病類別之判定,亦可為疾病類別之可能性。 Figure 3B is a schematic diagram of classifier A performing classification according to an embodiment of the present invention. As shown in Figure 3A, the image features used by classifier A are the results (ie, variability and image intensity) produced by the variability algorithm and the image intensity algorithm. In this embodiment, the
根據本發明之一實施例,運算模組230所產生之分析結果可包括對應第二影像之口腔得到口腔癌之機率。舉例來說,若以第3A-3B圖為例,若權重決定器310產生 之對應第二影像之分類器A-C之權重值W 1、W 2和W M分別為0.2、0.5和0.3,且經過分類器運算演算法之分類器A-C判定之結果是癌症之可能性為0.8,0.9,0.7,運算模組230透過加權評估後將可以得到對應第一影像之口腔罹患癌症的機率f en 是0.82(82%)。此外,運算模組230所產生之分析結果亦可包括,根據不同口腔癌之機率所對應之分析之以及建議。舉例來說,機率為0~10%可設定為正常,10~50%可設定為可能有異常建議使用者長期觀察,50~90%可設定為有異常建議使用者近期安排檢查,以及90~100%可設定為明顯異常為高風險,建議使用者盡快就醫檢查尋求醫師協助。 According to an embodiment of the present invention, the analysis result generated by the
第4A圖係根據本發明之一實施例所述之第二平台140之方塊圖。如第4A圖所示,第二平台140可包括一處理裝置141、一儲存裝置142,以及一顯示裝置143。需注意的是,在第4A圖中之方塊圖,僅係為了方便說明本發明之實施例,但本發明並不以此為限。第二平台140亦可包括其他元件。 FIG. 4A is a block diagram of the
根據本發明之一實施例,處理裝置141可係一包含一或複數處理器之裝置。根據本發明之一實施例,第二平台140可直接經由第一平台110或經由雲端網路130從第一平台110取得第一影像、複數分類器以及複數權重值,並將取得之第一影像、複數分類器以及複數權重值儲存在儲存裝置142。根據本發明另一實施例,第二平台140僅須從第一平台110取得第一影像,複數分類器和複數權 重值則可由第二平台產生,再傳給第一平台。在此實施例中,從第一平台110取得第一影像,以及第二平台產生之複數分類器和複數權重值產生亦會儲存在儲存裝置142。根據本發明之一實施例,顯示裝置143可為一液晶螢幕或一觸控面板,但本發明不以此為限。 According to an embodiment of the present invention, the
第4B圖係根據本發明之一實施例所述之處理裝置141之方塊圖。如第4圖所示,處理裝置141可包括一部位判定模組410、一選取模組420、一運算模組430以及一優化模組440。 FIG. 4B is a block diagram of the
根據本發明之實施例,第一平台110所進行之口腔影像分析方法亦可應用在第二平台140。此外,處理裝置141之部位判定模組410、選取模組420,以及運算模組430之操作類似部位判定模組210、選取模組220,以及運算模組230。因此,在此就不再贅述。 According to an embodiment of the present invention, the oral image analysis method performed by the
如第4圖所示,和處理裝置121相比,處理裝置141更包括一優化模組440。優化模組440可根據運算模組430所產生之分析結果,更新第一影像對應之部位所對應之分類器以及權重值,並將更新之資訊儲存到儲存裝置142。此外,第二平台140可經由雲端網路130將更新之資訊提供給第一平台120。 As shown in FIG. 4, compared with the
第5圖係根據本揭露之一實施例所述之口腔影像分析方法之流程圖。此口腔影像分析方法可適用本揭露之口腔影像分析系統100。在步驟S510,分析系統100之一影像擷取裝置會產生一第一影像。在步驟S520,分析系 統100之一平台(一第一平台或一第二平台,且第二平台之運算能力會優於第一平台之運算能力)會取得第一影像。在步驟S530,上述平台會判斷第一影像對應到口腔之一位置。在步驟S540,根據上述位置,上述平台會取得複數分類器和取得對應上述複數分類器之複數權重值。在步驟S550,上述平台會從第一影像中選取對應一可能病變區域之一第二影像。S560,上述平台會根據上述複數分類器、上述複數權重值以及上述第二影像產生一分析結果。 FIG. 5 is a flowchart of the oral image analysis method according to an embodiment of the disclosure. This oral image analysis method can be applied to the oral
根據本發明一些實施例,在口腔影像分析方法中,分析系統100之第二平台會透過一雲端網路從分析系統100之第一平台或直接從分析系統100之第一平台取得第一影像。第二平台會判斷第一影像對應到口腔之位置,且根據上述位置,取得複數分類器和取得對應複數分類器之複數權重值。此外,第二平台會從第一影像中選取對應可能病變區域之第二影像,以及根據上述複數分類器、上述複數權重值以及上述第二影像產生上述分析結果。 According to some embodiments of the present invention, in the oral image analysis method, the second platform of the
根據本發明一些實施例,在口腔影像分析方法中,第二平台會根據上述分析結果,更新上述複數分類器以及上述複數權重值。此外,根據本發明一些實施例,在口腔影像分析方法中,第二平台會透過雲端網路將更新之上述複數分類器以及上述複數權重值提供給第一平台。 According to some embodiments of the present invention, in the oral image analysis method, the second platform will update the complex classifier and the complex weight value according to the analysis result. In addition, according to some embodiments of the present invention, in the oral image analysis method, the second platform provides the updated complex classifier and the complex weight value to the first platform via the cloud network.
根據本發明之實施例所提出之口腔影像系統和方法,將可降低因人為操作上的差異所導致的診斷誤差,以及輔助醫生或研究人員對於切片位置之判定。此外,根 據本發明之實施例所提出之口腔影像系統和方法,醫生或研究人員可直接透過雲端之資料庫平台,根據資料庫平台之分析結果,進行病症之研究或討論。 According to the oral imaging system and method proposed by the embodiments of the present invention, the diagnosis error caused by the difference in human operation can be reduced, and the diagnosis of the slice position can be assisted by doctors or researchers. In addition, according to the oral imaging system and method proposed in the embodiments of the present invention, doctors or researchers can directly use the cloud database platform to conduct research or discussion of diseases based on the analysis results of the database platform.
在本說明書中以及申請專利範圍中的序號,例如「第一」、「第二」等等,僅係為了方便說明,彼此之間並沒有順序上的先後關係。 The serial numbers in this specification and in the scope of the patent application, such as "first", "second", etc., are only for convenience of description, and there is no sequential relationship between them.
本發明之說明書所揭露之方法和演算法之步驟,可直接透過執行一處理器直接應用在硬體以及軟體模組或兩者之結合上。一軟體模組(包括執行指令和相關數據)和其它數據可儲存在數據記憶體中,像是隨機存取記憶體(RAM)、快閃記憶體(flash memory)、唯讀記憶體(ROM)、可抹除可規化唯讀記憶體(EPROM)、電子可抹除可規劃唯讀記憶體(EEPROM)、暫存器、硬碟、可攜式硬碟、光碟唯讀記憶體(CD-ROM)、DVD或在此領域習之技術中任何其它電腦可讀取之儲存媒體格式。一儲存媒體可耦接至一機器裝置,舉例來說,像是電腦/處理器(為了說明之方便,在本說明書以處理器來表示),上述處理器可透過來讀取資訊(像是程式碼),以及寫入資訊至儲存媒體。一儲存媒體可整合一處理器。一特殊應用積體電路(ASIC)包括處理器和儲存媒體。一用戶設備則包括一特殊應用積體電路。換句話說,處理器和儲存媒體以不直接連接用戶設備的方式,包含於用戶設備中。此外,在一些實施例中,任何適合電腦程序之產品包括可讀取之儲存媒體,其中可讀取之儲存媒體包括和一或多個所揭露實施例相關之程式碼。在一些 實施例中,電腦程序之產品可包括封裝材料。 The method and algorithm steps disclosed in the specification of the present invention can be directly applied to hardware and software modules or a combination of the two directly by executing a processor. A software module (including execution commands and related data) and other data can be stored in data memory, such as random access memory (RAM), flash memory (flash memory), and read-only memory (ROM) , Erasable programmable read-only memory (EPROM), electronically erasable programmable read-only memory (EEPROM), scratchpad, hard disk, portable hard disk, optical disk read-only memory (CD- ROM), DVD, or any other computer-readable storage media format used in this field. A storage medium can be coupled to a machine device, such as a computer/processor (for the convenience of description, it is represented by a processor in this manual), and the processor can read information (such as a program Code), and write information to the storage medium. A storage medium can integrate a processor. An application-specific integrated circuit (ASIC) includes a processor and a storage medium. A user equipment includes a special application integrated circuit. In other words, the processor and the storage medium are included in the user equipment in a manner that is not directly connected to the user equipment. In addition, in some embodiments, any product suitable for computer programs includes a readable storage medium, where the readable storage medium includes code related to one or more of the disclosed embodiments. In some embodiments, the product of the computer program may include packaging materials.
以上段落使用多種層面描述。顯然的,本文的教示可以多種方式實現,而在範例中揭露之任何特定架構或功能僅為一代表性之狀況。根據本文之教示,任何熟知此技藝之人士應理解在本文揭露之各層面可獨立實作或兩種以上之層面可以合併實作。 The above paragraphs use multiple levels of description. Obviously, the teachings of this document can be implemented in various ways, and any specific structure or function disclosed in the example is only a representative situation. According to the teachings of this article, anyone who is familiar with this technique should understand that each level disclosed in this article can be implemented independently or two or more levels can be combined.
雖然本揭露已以實施例揭露如上,然其並非用以限定本揭露,任何熟習此技藝者,在不脫離本揭露之精神和範圍內,當可作些許之更動與潤飾,因此發明之保護範圍當視後附之申請專利範圍所界定者為準。 Although this disclosure has been disclosed in the above embodiments, it is not intended to limit the disclosure. Anyone who is familiar with this technique can make some changes and modifications without departing from the spirit and scope of this disclosure. Therefore, the protection scope of the invention The scope of the patent application attached hereafter shall prevail.
100‧‧‧口腔影像分析系統 100‧‧‧Oral Image Analysis System
110‧‧‧影像擷取裝置 110‧‧‧Image capture device
120‧‧‧第一平台 120‧‧‧First Platform
130‧‧‧雲端網路 130‧‧‧Cloud Network
140‧‧‧第二平台 140‧‧‧Second Platform
Claims (20)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW107123466A TWI703961B (en) | 2018-07-06 | 2018-07-06 | Oral image analysis system and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW107123466A TWI703961B (en) | 2018-07-06 | 2018-07-06 | Oral image analysis system and method |
Publications (2)
Publication Number | Publication Date |
---|---|
TW202005609A TW202005609A (en) | 2020-02-01 |
TWI703961B true TWI703961B (en) | 2020-09-11 |
Family
ID=70412930
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW107123466A TWI703961B (en) | 2018-07-06 | 2018-07-06 | Oral image analysis system and method |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI703961B (en) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI784334B (en) | 2020-10-29 | 2022-11-21 | 國立臺灣大學 | Disease diagnosing method and disease diagnosing system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20060122516A (en) * | 2005-05-27 | 2006-11-30 | (주)누리젠 | Diagnosis system of cancer of the mouth using a internet |
CN107735838A (en) * | 2017-03-02 | 2018-02-23 | 香港应用科技研究院有限公司 | It is used for the abnormality detection of medical sample under a variety of settings |
CN108062575A (en) * | 2018-01-03 | 2018-05-22 | 广东电子工业研究院有限公司 | A kind of high similarity graph picture identification and sorting technique |
-
2018
- 2018-07-06 TW TW107123466A patent/TWI703961B/en active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR20060122516A (en) * | 2005-05-27 | 2006-11-30 | (주)누리젠 | Diagnosis system of cancer of the mouth using a internet |
CN107735838A (en) * | 2017-03-02 | 2018-02-23 | 香港应用科技研究院有限公司 | It is used for the abnormality detection of medical sample under a variety of settings |
CN108062575A (en) * | 2018-01-03 | 2018-05-22 | 广东电子工业研究院有限公司 | A kind of high similarity graph picture identification and sorting technique |
Also Published As
Publication number | Publication date |
---|---|
TW202005609A (en) | 2020-02-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2018201632A1 (en) | Artificial neural network and system for recognizing lesion in fundus image | |
CN112088394A (en) | Computerized classification of biological tissue | |
WO2022007337A1 (en) | Tumor cell content evaluation method and system, and computer device and storage medium | |
Kauppi | Eye fundus image analysis for automatic detection of diabetic retinopathy | |
KR20200116107A (en) | Automated monitoring of medical imaging procedures | |
CN109948671B (en) | Image classification method, device, storage medium and endoscopic imaging equipment | |
CN112037913B (en) | Periodontitis intelligent detection system based on convolutional neural network | |
RU2633320C2 (en) | Selection of images for optical study of uterine cervix | |
EP2344982A1 (en) | Methods for tissue classification in cervical imagery | |
JP2015500722A (en) | Method and apparatus for detecting and quantifying skin symptoms in a skin zone | |
CN113962311A (en) | Knowledge data and artificial intelligence driven ophthalmic multi-disease identification system | |
JP2013200590A (en) | Similar image retrieval device, method, and program | |
CN111369501B (en) | Deep learning method for identifying oral squamous cell carcinoma based on visual features | |
CN113243887B (en) | Intelligent diagnosis and treatment instrument for macular degeneration of old people | |
Tobias et al. | CNN-based deep learning model for chest X-ray health classification using tensorflow | |
ALbahbah et al. | Detection of caries in panoramic dental X-ray images using back-propagation neural network | |
CN112927179A (en) | Intelligent analysis method for liver tumor | |
Kendrick et al. | Translating clinical delineation of diabetic foot ulcers into machine interpretable segmentation | |
CN113782184A (en) | Cerebral apoplexy auxiliary evaluation system based on facial key point and feature pre-learning | |
CN107066779A (en) | Self-service body parameter inspection method and system based on biological specimen color | |
Ay et al. | Automated classification of nasal polyps in endoscopy video-frames using handcrafted and CNN features | |
TWI703961B (en) | Oral image analysis system and method | |
Xu et al. | Artificial intelligence assisted identification of therapy history from periapical films for dental root canal | |
CN114332910A (en) | Human body part segmentation method for similar feature calculation of far infrared image | |
Vimalarani et al. | Automatic diagnosis and detection of dental caries in bitewing radiographs using pervasive deep gradient based LeNet classifier model |